The stack

Run the whole business
on open source.

405 tools across 68 layers — every function a company actually has, what each tool is in one line, its licence, and the paid product it replaces. 55 of them are already on this machine.

live running today here installed, not wired red licence = copyleft or restricted: run it, don't resell it

ai

The AI core

The engine room. Nothing here faces a customer — it is what lets every other layer think, remember and act instead of just storing rows.

Inference servers

The thing that actually runs a model's weights and answers requests. Pick by where it runs: GPU server, laptop CPU, or a phone.

vLLM
The production standard for serving open models on GPUs — paged attention, high throughput
Apache-2.0
OpenAI API spend
SGLang
vLLM's faster rival on structured output and multi-turn agent traffic
Apache-2.0
OpenAI API spend
llama.cpplive
Runs quantised models on plain CPU/consumer GPU — the reason a laptop can host an LLM
MIT
any hosted API
Ollamalive
One-command wrapper over llama.cpp — `ollama run`, model registry, local API
MIT
LocalAI
Drop-in OpenAI-compatible server for text, image, audio and embeddings in one binary
MIT
OpenAI API
TGI (Text Generation Inference)
HuggingFace's serving stack, tight with the Hub
Apache-2.0
hosted inference
LMDeploy
Very fast quantised serving out of the InternLM group
Apache-2.0
KTransformers
Runs huge MoE models on one GPU + system RAM
Apache-2.0
8×A100 rental
Jan
Desktop app — ChatGPT-shaped UI over local models, no account
AGPL-3.0
ChatGPT Plus $20/mo
colibri
Runs GLM-5.2 (744B MoE) on a 25GB-RAM box with no GPU
MIT
frontier API spend

Model gateways & routers

One endpoint in front of every model you use. Handles keys, failover, cost tracking and per-request routing so your apps never hardcode a provider.

LiteLLM
The default — 100+ providers behind one OpenAI-shaped API, with budgets and keys
MIT
per-provider SDK sprawl
OmniRoutelive
Ours. Groq + OpenRouter + Gemini + local, running on :20128
internal
Bifrost
Go gateway built for very high throughput and low overhead
Apache-2.0
RouteLLM
Sends easy prompts to a cheap model, hard ones to a strong one
Apache-2.0
50-80% of token spend
optillm
Inference proxy that applies reasoning tricks to raise answer quality
Apache-2.0
cascadeflow
Cascading runtime — try small model first, escalate only on failure
Apache-2.0
token spend
Portkey Gateway
Gateway with guardrails, caching and observability baked in
Apache-2.0

Agent frameworks

Libraries that let a model plan, call tools and loop until a job is done. Code-first ones give control; visual ones give speed.

LangGraph
Graph-shaped agent runtime with real state, retries and human-in-the-loop
MIT
CrewAI
Role-based multi-agent crews — researcher, writer, reviewer
MIT
AG2 / AutoGen
Microsoft's conversational multi-agent framework
Apache-2.0
Pydantic AI
Typed, validated agent outputs — the sane choice for production Python
MIT
smolagents
HuggingFace's tiny agent lib — agents that write and run code
Apache-2.0
Agno
Fast multi-modal agents with memory and knowledge built in
MPL-2.0
Mastra
The TypeScript answer — agents, workflows and RAG for Node teams
Apache-2.0
Letta (MemGPT)here
Stateful agents that keep long-term memory across sessions
Apache-2.0
Dify
Visual LLM app builder — RAG, agents, datasets, publishable APIs
Apache-2.0*
OpenAI Assistants
Flowise
Drag-and-drop LangChain — fastest way to a working chatbot
Apache-2.0
Voiceflow $50/mo
Langflow
Same idea, cleaner UI, strong on RAG pipelines
MIT
OpenHands
Autonomous software engineer — reads a repo, writes code, opens PRs
MIT
Devin $500/mo

Coding agents

Agents pointed at a repo instead of a chat box. This is the layer that actually replaces contract dev hours.

Claude Codelive
What this operation runs on — terminal agent with full tool access
proprietary
Aider
Terminal pair-programmer that commits its own work to git
Apache-2.0
Copilot $10/mo
Cline
VS Code agent — plan/act loop, MCP support, browser use
Apache-2.0
Cursor $20/mo
Roo Code / Kilo Code
Cline forks with multi-mode agents and cheaper routing
Apache-2.0
Cursor $20/mo
Continue
Open IDE autocomplete + chat, bring your own model
Apache-2.0
Copilot $10/mo
Goose
Block's on-machine agent — runs shell, edits files, extensible
Apache-2.0
OpenCode
Terminal coding agent, provider-agnostic
MIT
SWE-agent
Research agent that fixes real GitHub issues end to end
MIT

Vector databases

Where embeddings live so a model can find the right 20 paragraphs out of a million. If you already run Postgres, start with pgvector and stop there.

pgvector
Vector search inside Postgres — no new database to operate
PostgreSQL
Pinecone $70/mo
Qdrant
Rust vector DB, excellent filtering, easy self-host
Apache-2.0
Pinecone $70/mo
Milvushere
Scales to billions of vectors; heavier to run
Apache-2.0
Pinecone
Weaviate
Vector DB with built-in modules for embedding and reranking
BSD-3
Pinecone
Chroma
The prototyping default — pip install and go
Apache-2.0
LanceDB
Embedded, file-based vector DB — SQLite of vectors
Apache-2.0
ParadeDB
Postgres with real full-text + vector search bolted in
AGPL-3.0
Elastic

RAG frameworks

Retrieval-augmented generation: answer from your documents, with citations, instead of from the model's memory. This is 80% of real business AI.

LlamaIndex
The most complete RAG toolkit — loaders, indexes, query engines
MIT
Haystack
deepset's production RAG/search pipelines
Apache-2.0
RAGFlow
RAG engine with deep document understanding — tables and layout survive
Apache-2.0
R2R
RAG as a service you host — ingestion, graph, auth, API
MIT
Onyx (Danswer)
Company-wide AI search over Slack, Drive, Confluence, email
MIT
Glean $30/user/mo
Khoj
Personal AI over your notes, docs and email
AGPL-3.0
DocsGPT
Private assistant over a documentation set
MIT
open-notebookhere
Open NotebookLM — notebooks over your sources
MIT
NotebookLM

Agent memory

The difference between a chatbot and an assistant. Persists facts, preferences and history across sessions instead of starting cold.

gmemlive
Ours. Scoped memory engine wired into every CLI session
internal
Mem0
Drop-in memory layer — extracts and recalls facts per user
Apache-2.0
Zep
Temporal knowledge graph memory for agents
Apache-2.0
Graphiti
Builds a time-aware graph from conversation — Zep's engine
Apache-2.0
Cognee
Turns documents and chats into a queryable memory graph
Apache-2.0
memsearch
Unified persistent memory layer across agents
Apache-2.0

Fine-tuning

Teaching a base model your domain, your tone or your format. LoRA on a single GPU is enough for almost every business case.

Unsloth
2× faster LoRA on one consumer GPU — the practical default
Apache-2.0
GPU hours
Axolotl
Config-file fine-tuning across most open models
Apache-2.0
LLaMA-Factory
Fine-tune 100+ models from a UI or CLI
Apache-2.0
TRL + PEFT
HuggingFace's SFT/DPO/LoRA building blocks
Apache-2.0
torchtune
PyTorch-native recipes, no abstraction tax
BSD-3

Evals & observability

Proof that a change made things better, and a trace when a customer says it got something wrong. Skipping this is why AI features quietly rot.

Langfuse
The standard — traces, costs, prompt versions, evals, self-hosted
MIT
LangSmith $39/user/mo
promptfoolive
Prompt regression tests you run in CI, plus red-teaming
MIT
DeepEval
Pytest for LLMs — assert on faithfulness, relevance, bias
Apache-2.0
Ragas
The metric set for grading RAG answers
Apache-2.0
Phoenix (Arize)
Local tracing and eval UI, strong on embedding drift
ELv2
Opik
Comet's tracing + eval stack
Apache-2.0
OpenLLMetry
OpenTelemetry for LLM calls — plugs into Grafana
Apache-2.0
Helicone
One-line proxy that logs every call and its cost
Apache-2.0

Guardrails & AI security

Stops prompt injection, PII leaks and jailbreaks before they reach a customer or a payer. Non-negotiable once real data flows.

NeMo Guardrails
NVIDIA's rails — define allowed topics and flows in config
Apache-2.0
Guardrails AI
Validators on model output — types, PII, toxicity, format
Apache-2.0
LLM Guard
Scans prompts and responses for injection, secrets and PII
MIT
garak
NVIDIA's LLM vulnerability scanner — attacks your own model
Apache-2.0
pentest engagement
Rebuff
Prompt-injection detection with a canary-token trick
Apache-2.0
Giskard
Automated scanning for bias, robustness and performance holes
Apache-2.0

Chat interfaces

The window people actually type into. Put one in front of your gateway and the whole company gets ChatGPT without the per-seat bill.

Open WebUIhere
The most complete self-hosted chat UI — RAG, users, tools. ⚠ branding clause past 50 users
custom
ChatGPT Team $25/user/mo
LibreChat
Multi-provider chat with agents, files, code interpreter
MIT
ChatGPT Team
Lobe Chat
Polished UI, plugin market, artifacts
Apache-2.0*
AnythingLLM
Chat + RAG workspaces, desktop or docker
MIT
Cherry Studio
Desktop client across every provider
Apache-2.0

MCP & tool plumbing

The standard that lets one agent reach your email, DB, CRM and browser without a bespoke integration each time.

FastMCPlive
Build an MCP server in a few lines of Python
Apache-2.0
MCP servers (official)live
Reference servers — filesystem, git, postgres, fetch, more
MIT
Composio
250+ pre-built tool integrations for agents with auth handled
Elastic-2.0
Zapier $30/mo
mcp-agent
Agent patterns built on MCP primitives
Apache-2.0
media

Voice, vision & media

Everything that turns text into something a human hears or sees, and back again. This is the layer that lets an agent answer a phone or cut an ad.

Text to speech

Turns text into a voice. The split that matters: CPU-only models are free and instant; GPU models sound better and can clone a voice.

Kokoro
82M params, CPU-only, startlingly good — the current default for most uses
Apache-2.0
ElevenLabs $22/mo
pocket-ttshere
Kyutai's CPU TTS with cloning built in — tested here at 1.48x realtime
MIT
ElevenLabs
Piperhere
Tiny ONNX voices, near-instant, huge language catalogue
MIT
ElevenLabs
F5-TTS
Zero-shot cloning at high quality. Cloning a real person needs written consent
MIT
ElevenLabs Pro
Chatterbox
Resemble's open model — emotion control, very natural
MIT
ElevenLabs
Fish Speech
Strong multilingual cloning, excellent Chinese and Japanese
Apache-2.0
Orpheus TTS
Llama-based speech with laughs, sighs and real prosody
Apache-2.0
XTTS / Coqui
The long-standing cloning workhorse, 17 languages
MPL-2.0
vexyl-tts
22 Indian languages, 44 speakers — the Malayalam and Tamil unblock
Apache-2.0
Sarvam per-char
MeloTTS / OpenVoice
Fast multilingual, instant cloning from a short clip
MIT

Speech to text

Turns audio into text. Whisper is the family; the forks differ on speed, word timestamps and knowing who was speaking.

faster-whisperhere
CTranslate2 rewrite — 4x faster, same accuracy. The practical default
MIT
AssemblyAI $0.37/hr
whisper.cpp
C++ port that runs on a laptop, a Pi, even a phone
MIT
WhisperX
Whisper plus word-level timestamps and speaker labels
BSD-4
Rev $1.50/min
Parakeet (NeMo)
NVIDIA's model — currently the fastest accurate English ASR
CC-BY-4.0
Vosk
Fully offline, tiny models, runs on Android and Raspberry Pi
Apache-2.0
Moonshine
Built for real-time on-device transcription
MIT
SenseVoice / FunASR
Strong on Asian languages, plus emotion and event tags
MIT
Speaches
OpenAI-compatible STT and TTS server you self-host
MIT
OpenAI audio API

Realtime voice agents

The full loop — listen, think, speak, handle interruption — under 800ms. What a sales or support call agent actually needs.

LiveKit Agents
Strongest option — WebRTC, telephony, turn detection, production-proven
Apache-2.0
Vapi $0.05/min
Pipecathere
Composable voice pipeline, provider-agnostic, easy to reason about
BSD-2
Vapi
Vocode
Python voice agents wired to Twilio out of the box
MIT
Bland $0.09/min
dograhhere
Self-hosted Vapi alternative with a UI and call analytics
Apache-2.0
Vapi
Ultravox
Speech-in speech-out LLM — skips transcription entirely
MIT
OpenAI Realtime
Moshi
Full-duplex speech model — it can be interrupted mid-sentence
Apache-2.0

Diarisation & VAD

Who spoke, and when. Without this a meeting transcript is one wall of text with no names on it.

pyannote.audio
The standard speaker-diarisation toolkit
MIT
Silero VAD
Tiny, fast voice-activity detection — knows when speech starts
MIT
NeMo diarisation
NVIDIA's pipeline, strong on overlapping speech
Apache-2.0

Image generation

Marketing graphics, product shots, thumbnails, logos. ComfyUI is GPL — fine to run, do not bundle it inside something you sell closed.

ComfyUI
Node-graph image and video pipeline — the pro standard. Copyleft
GPL-3.0
Midjourney $30/mo
FLUX.1 dev
Best open image quality. The dev weights are non-commercial
non-commercial
Midjourney
Stable Diffusion XL
The workhorse — fully commercial, huge LoRA ecosystem
OpenRAIL++
Midjourney
Fooocus
One-box UI that hides every knob — closest to the Midjourney feel
GPL-3.0
Midjourney
InvokeAI
Studio UI with real canvas and inpainting workflow
Apache-2.0
Photoshop AI
Krita + AI plugin
Full painting app with generative fill
GPL-3.0
Photoshop $23/mo

Video generation & editing

Ads, product motion, subtitled clips. Generation needs a GPU; the editing and clipping tools are CPU and free.

LTX-Videolive
Fast text-to-video. Already wired here through GitHub Actions
Apache-2.0
Runway $15/mo
Wan 2.x
Alibaba's video model — strong motion and text rendering
Apache-2.0
Runway
HunyuanVideo
Tencent's large video model, high quality
custom
FramePack
Long video on 6GB VRAM by packing frames
Apache-2.0
FunClip
Transcribe, subtitle and LLM-clip long video into shorts
MIT
Opus Clip $29/mo
MoviePy / ffmpeglive
Programmatic editing — the backbone of every render pipeline
MIT / LGPL
auto-editor
Strips silence and dead air automatically
public domain
Descript $24/mo

Avatars & talking heads

A face that speaks your script. The weakest part of open source — usable for internal video, not for a client's brand film.

LatentSync
ByteDance lip-sync — currently the best open result
Apache-2.0
HeyGen $29/mo
SadTalker
One photo plus audio becomes a talking head. Rough but free
Apache-2.0
HeyGen
MuseTalk
Realtime lip-sync at 30fps
MIT
EchoMimic
Audio-driven portrait animation with better expression
Apache-2.0
OpenAvatarChat
Full realtime avatar conversation stack
Apache-2.0

Document AI & OCR

Turning PDFs, scans and faxes into structured data. For RCM this is the front door — EOBs, remits, payer manuals, contracts.

Doclinglive
IBM's parser — layout, tables and reading order preserved. Best default
MIT
Docparser $99/mo
MinerU
Complex PDF to markdown with tables intact. Read the licence before client work
AGPL-3.0
Azure Doc Intelligence
Surya
OCR plus layout and reading order in 90+ languages
GPL-3.0
olmOCR
AllenAI's vision-model OCR, very strong on messy scans
Apache-2.0
PaddleOCR
Battle-tested multilingual OCR with table recognition
Apache-2.0
Tesseract
The classic engine — fast, free, weaker on layout
Apache-2.0
Marker
PDF to clean markdown, fast
GPL-3.0
Unstructured
Ingest any file type into chunks ready for RAG
Apache-2.0
markpdfdown
LLM-assisted PDF to markdown, CPU only
MIT

Vision & detection

Counting, reading and spotting things in images or video. Ultralytics YOLO is AGPL and enforced — in a product you sell it means publishing your source.

YOLO (Ultralytics)
Fastest path to object detection. The licence bites on resale
AGPL-3.0
RT-DETR
Transformer detector, commercially safe alternative to YOLO
Apache-2.0
SAM 2
Meta's segment-anything — click a thing, get its exact mask
Apache-2.0
Grounding DINO
Detect objects by describing them in words
Apache-2.0
Supervision
The glue — annotate, track and count over any detector
MIT
gtm

Go to market

Finding people, reaching them, and closing them. Every row here has a SaaS bill attached to it in most companies — this is where self-hosting pays back fastest.

CRM & system of record

The one place a customer's history lives. Getting this wrong means every other tool argues about who the customer is.

Stormlive
Ours — 45K leads on MongoDB, live at storm.getmaxglobal.com
internal
Salesforce
Twentyhere
The modern open Salesforce alternative, built for AI. Cloned here
AGPL-3.0
Salesforce $165/user/mo
EspoCRM
Mature, boring, works — full sales and support CRM
AGPL-3.0
HubSpot $90/user/mo
SuiteCRM
The SugarCRM fork — heavyweight, deeply customisable
AGPL-3.0
Salesforce
Odoo CRM
CRM as one module of a full ERP — good if you want one system
LGPL-3.0
HubSpot
Krayin
Laravel CRM, light and easy to modify
MIT
Relaticlehere
CRM with native AI agent support and 30 MCP tools
AGPL-3.0

Lead generation & enrichment

Filling the top of the funnel legally. Scraping LinkedIn is fragile and bannable — local business data and public registries are not.

google-maps-scraperhere
Local business leads with phone, site and rating. Cloned here
MIT
Apollo $99/mo
Firecrawl
Turns any site into clean markdown for an LLM to read
AGPL-3.0
Crawl4AI
Async crawler built specifically to feed LLMs
Apache-2.0
Scrapy
The industrial Python crawler framework
BSD-3
Katana
Fast site crawler from ProjectDiscovery
MIT
Reacher
Self-hosted email verification — is this address real
AGPL-3.0
ZeroBounce $0.008/email
NPPES / NPI registrylive
Free US provider registry — the RCM lead source
public
CAQH fees

Marketing automation & email campaigns

Sequences, newsletters, drip campaigns. Host the engine, but always send through a real provider — port 25 from your own box will not deliver.

Listmonk
Single Go binary, millions of emails, very fast. The best pick
AGPL-3.0
Mailchimp $50/mo
Mautic
Full marketing automation — scoring, journeys, landing pages
GPL-3.0
HubSpot $800/mo
Dittofeed
Developer-first customer journeys and messaging
MIT
Braze
Keila
Simple, pretty newsletter tool
AGPL-3.0
ConvertKit $29/mo
Postal
Your own full mail server for outbound at volume
MIT
SendGrid $20/mo
useSend
Open Resend alternative — transactional email API
AGPL-3.0
Resend $20/mo
docker-mailserver
Production mailserver in a container — SMTP, IMAP, spam
MIT
Google Workspace

Social publishing

Scheduling and posting across platforms. The tool is never the blocker — getting each platform's API app approved is the real work.

Postiz
Modern scheduler, 15+ platforms, AI captions
AGPL-3.0
Buffer $6/channel/mo
Mixpost
Self-hosted scheduling, one-time licence, no per-channel fee
custom
Hootsuite $99/mo
social-enginehere
Ours — the LinkedIn queue and posting scripts
internal

SEO & content intelligence

Keywords, rankings, crawl health, trends. This whole layer is CPU-only and free — there is no reason to be paying Semrush.

open-seo
Open alternative to Semrush and Ahrefs — keywords, competitors, audits
MIT
Semrush $139/mo
crawlseo
SEO monitoring — Search Console, crawler, Core Web Vitals in one dashboard
MIT
Ahrefs $99/mo
trends-checker
Google Trends from the command line
MIT
Glimpse $49/mo
SerpBear
Track your keyword positions over time
MIT
AccuRanker $116/mo
Morphic
AI search engine with generative UI — research assistant
Apache-2.0
Perplexity Pro $20/mo

Web analytics & product analytics

Who came, what they did, where they dropped. Self-hosting also solves the consent problem — the data never leaves your box.

PostHog
Product analytics, session replay, flags and A/B in one. The big one
MIT*
Amplitude $61k/yr
Plausible
Light, privacy-first, no cookie banner needed
AGPL-3.0
Google Analytics
Umami
Tiny, fast, beautiful web analytics
MIT
GA / Fathom $14/mo
Matomo
The full GA replacement with heatmaps and funnels
GPL-3.0
GA360 $150k/yr
Rybbit
New generation, very clean UI
AGPL-3.0

Experiments & feature flags

Ship a change to 5% of users, measure, then roll out. Also the safest way to launch AI features you are not sure about.

Unleash
The mature open flag service
Apache-2.0
LaunchDarkly $10/seat
GrowthBook
Flags plus a real stats engine for A/B tests
MIT*
Optimizely $36k/yr
Flagsmith
Flags, remote config, segments
BSD-3
LaunchDarkly
OpenFeature
The vendor-neutral SDK standard — write once, swap backends
Apache-2.0

Website, CMS & landing pages

Where the content and the sign-up form live. Headless CMS keeps the front-end yours and the editing simple for non-developers.

Strapi
The most-used headless CMS — content types, roles, REST and GraphQL
MIT*
Contentful $300/mo
Directus
Turns any SQL database into a CMS and admin panel
BSL
Contentful
Payload
TypeScript-native CMS that lives inside your Next.js app
MIT
Sanity $99/mo
Ghost
Publishing and paid newsletters in one
MIT
Substack 10% cut
Astro
Static-first framework — the fastest way to a fast marketing site
MIT
Webflow $23/mo
Webstudio
Visual builder that outputs real code, not a lock-in
AGPL-3.0
Webflow

Ecommerce & storefronts

Selling product online. Medusa is the developer pick; Bagisto and WooCommerce win when a non-developer has to run the shop.

Medusa
Headless commerce in TypeScript — the modern default
MIT
Shopify $79/mo + 2%
Saleor
GraphQL-native, enterprise scale
BSD-3
Shopify Plus $2000/mo
Vendure
TypeScript commerce with a strong plugin model
GPL-3.0
Shopify
Bagisto
Laravel shop with a full admin — good for handing to a client
MIT
Shopify
WooCommerce
The WordPress route — biggest ecosystem, easiest to hire for
GPL-3.0
Shopify
Paymenter
Webshop and billing for hosting or service products
MIT

Booking & scheduling

Letting someone pick a time without an email thread. One of the highest value-per-hour things to self-host.

Cal.com
The open Calendly — routing, teams, payments, API
AGPL-3.0
Calendly $12/user/mo
Easy!Appointments
Simple service-business booking with a calendar
GPL-3.0
Acuity $16/mo
Rallly
Group poll for finding a time — the Doodle replacement
AGPL-3.0
Doodle $6.95/mo

Forms & surveys

Intake, qualification, feedback. Formbricks is the one that also does the in-app survey side.

Formbricks
Surveys and in-product feedback with targeting
AGPL-3.0
Typeform $25/mo
Typebot
Conversational forms — chat-style intake that converts better
AGPL-3.0
Typeform
SurveyJS
Form engine and visual builder you embed in your own app
MIT
Jotform $34/mo
ops

Run the work

Communication, projects, documents, meetings, people. The unglamorous layer where a 20-person company quietly spends $3-5k a month on seats.

Messaging & customer conversations

Reaching a person on the channel they actually read. In India that is WhatsApp, and self-hosting it is the single biggest unlock here.

Baileyslive
WhatsApp Web protocol in TypeScript — the engine behind our gateway
MIT
Twilio WA $0.005/msg
go-whatsapp-web-multidevice
WhatsApp REST API with a UI and multi-account support
MIT
Twilio WA
Chatwoot
Shared inbox across WhatsApp, email, chat and social. The best pick
MIT
Intercom $74/mo
textbeehere
Turns any spare Android phone into an SMS gateway. Cloned here
MIT
Twilio SMS
Typebot
Chat flows for qualification before a human picks up
AGPL-3.0
Novu
One API for email, SMS, push and in-app notifications
MIT
Knock $250/mo
ntfy / Gotify
Dead simple self-hosted push to your phone from a script
Apache-2.0 / MIT
Pushover

Helpdesk & support

Tickets, SLAs and a knowledge base. Only worth standing up once more than one person answers customers.

Zammad
Modern helpdesk — email, chat, telephony, good reporting
AGPL-3.0
Zendesk $55/agent/mo
FreeScout
Light Help Scout clone, very easy to run
AGPL-3.0
Help Scout $25/user/mo
osTicket
The old reliable — plain, stable, no surprises
GPL-2.0
Zendesk
Chaskiq
Support plus in-app messaging and campaigns
AGPL-3.0
Intercom

Team chat

Internal conversation. Matrix is the one that federates; Mattermost is the one enterprises approve.

Mattermost
The serious Slack replacement — channels, threads, compliance
AGPL-3.0*
Slack $8.75/user/mo
Rocket.Chathere
Chat plus omnichannel customer contact. Cloned here
MIT
Slack
Zulip
Threaded by topic — much better for async and searchable history
Apache-2.0
Slack
Element / Matrix
Federated and end-to-end encrypted — the protocol, not just an app
Apache-2.0
Slack

Meetings & video

Calls, webinars and the transcript afterwards. LiveKit is the building block; Jitsi is the finished product.

Jitsi Meet
Self-hosted video calls with no accounts and no time limit
Apache-2.0
Zoom $16/host/mo
LiveKit
WebRTC infrastructure — build your own calling into a product
Apache-2.0
Twilio Video
BigBlueButton
Built for teaching — whiteboard, breakouts, polls
LGPL-3.0
Zoom Webinar $79/mo
Lisalive
Ours — meeting ops and native Teams transcripts through Graph
internal
Otter $20/user/mo
Vexahere
Self-hosted bot that joins Meet and Teams and transcribes. Cloned
Apache-2.0
Fireflies $18/user/mo
Meetily
Privacy-first meeting notes, runs fully local
MIT
Otter
Hyprnote
Local-first AI notepad that listens to the meeting
GPL-3.0
Granola $18/mo

Project management & tickets

Where work is tracked. Plane is the closest open Linear; Huly is the everything-app bet.

Plane
Clean, fast, modern — issues, cycles, modules. The best open Jira
AGPL-3.0
Jira $8.60/user/mo
Hulyhere
Jira plus Notion plus Slack in one product. Cloned here
EPL-2.0
Jira + Notion
OpenProject
Classic PM with Gantt, budgets and formal reporting
GPL-3.0
MS Project $10/user/mo
Vikunja
Light task app — lists, kanban, gantt, no ceremony
AGPL-3.0
Todoist $4/user/mo
Focalboard
Self-hosted Trello and Notion boards
MIT
Trello $5/user/mo
Taiga
Agile-native — scrum and kanban done properly
MPL-2.0
Jira
Leantime
PM designed for non-project-managers and ADHD brains
AGPL-3.0

Docs, wiki & knowledge base

Where decisions and process live so they survive people leaving. This is also the corpus your RAG layer should read.

Outline
Fast team wiki with excellent search. Best-looking option
BSL-1.1
Notion $10/user/mo
Docmost
Modern collaborative wiki — real-time editing, spaces
AGPL-3.0
Confluence $6/user/mo
AFFiNE
Docs plus whiteboard plus database in one canvas
MIT
Notion + Miro
AppFlowy
Local-first Notion alternative with a desktop app
AGPL-3.0
Notion
BookStack
Shelves, books, chapters — the clearest structure for SOPs
MIT
Confluence
Wiki.js
Markdown wiki with git sync and many auth backends
AGPL-3.0
Confluence
Logseq / SiYuan
Personal knowledge graphs — outliner style, local files
AGPL-3.0
Roam $15/mo
getmax-brain-kblive
Ours — the 8-dimension KB in markdown and git
internal

Files, storage & sync

The company drive. Nextcloud is the full office suite; MinIO is the S3 your applications talk to.

Nextcloud
Drive, calendar, contacts, office, talk — the full Google Workspace swap
AGPL-3.0
Workspace $12/user/mo
Seafile
Faster and more reliable pure file sync than Nextcloud
AGPL-3.0
Dropbox $15/user/mo
Syncthing
Peer-to-peer sync with no server at all
MPL-2.0
Dropbox
MinIOhere
S3-compatible object storage you run yourself
AGPL-3.0
S3 storage cost
Garage
Lightweight S3 for small self-hosted setups
AGPL-3.0
S3
rclone
Move and mount data between 70+ storage backends
MIT

E-signature & documents

Contracts signed without DocuSign. Both good options are copyleft — fine for your own contracts, careful if you rebrand and resell.

Documensohere
The modern open DocuSign — clean UX, good API
AGPL-3.0
DocuSign $25/user/mo
DocuSealhere
Form-field heavy documents and templates. Cloned here
AGPL-3.0
DocuSign
OpenSign
Full e-sign with audit trail and templates
AGPL-3.0
DocuSign

Whiteboard & design

Diagrams, wireframes, architecture. Excalidraw is the one everyone actually uses; Penpot is the real Figma replacement.

Excalidraw
Hand-drawn style whiteboard — the fastest way to explain a system
MIT
Miro $8/user/mo
tldraw
Infinite canvas SDK you can embed in your own product
custom
Miro
Penpot
Design and prototyping that speaks real CSS — the Figma alternative
MPL-2.0
Figma $15/user/mo
draw.io
Formal diagrams, AWS icons, offline desktop app
Apache-2.0
Lucidchart $9/user/mo
Mermaidlive
Diagrams written as text in markdown — version-controllable
MIT

People, HR & time

Onboarding, leave, attendance, appraisals. Frappe HR is the only open option that covers Indian payroll properly.

Frappe HR
Full HRMS — leave, attendance, payroll, appraisals, India-ready
GPL-3.0
Keka ₹9000/mo
OrangeHRM
Long-standing HR suite, wide module coverage
GPL-2.0
BambooHR $250/mo
Horilla
Modern Django HRMS with recruitment built in
LGPL-2.1
Zoho People
Kimai
Time tracking and invoicing for billable work
AGPL-3.0
Toggl $9/user/mo
ActivityWatchhere
Automatic time tracking on the machine. Cloned here
MPL-2.0
RescueTime $12/mo
OpenCATS
Applicant tracking for recruiting
AGPL-3.0
Greenhouse $6.5k/yr

Learning & enablement

Training staff or selling a cohort. Moodle is the institution standard; Frappe LMS is the light one.

Moodle
The world's most-used LMS — courses, quizzes, certificates
GPL-3.0
Teachable $39/mo
Frappe LMS
Clean, modern, integrates with the Frappe stack
AGPL-3.0
Thinkific $49/mo
Open edX
University-scale course platform
AGPL-3.0
Build45live
Ours — the paid cohort product
internal
money

Money & back office

Invoices, books, payroll, compliance. The layer founders postpone until an auditor asks — and the one where copyleft licences matter most, because you run these for yourself rather than resell them.

ERP — the whole back office in one

If you would otherwise stitch six tools together, an ERP is usually the cheaper answer. ERPNext is the strongest open option for an Indian company.

ERPNext
Accounting, inventory, HR, CRM, projects, GST-ready. The best all-in-one
GPL-3.0
Zoho One ₹1500/user/mo
Odoo Community
Huge module ecosystem. The community edition is genuinely usable
LGPL-3.0
Odoo Enterprise $31/user/mo
Dolibarr
Light ERP and CRM for small businesses, easy to install
GPL-3.0
Axelor
Java ERP with a low-code studio for custom modules
AGPL-3.0
Tryton
Modular, strict, accountant-friendly
GPL-3.0

Accounting & bookkeeping

The books themselves. Plain-text accounting is the developer's answer; Akaunting and ERPNext are what an accountant will accept.

ERPNext Accounts
Double entry, GST, TDS, e-invoicing for India
GPL-3.0
Tally ₹18000/yr
Akaunting
Clean small-business accounting with an app store
GPL-3.0
QuickBooks $30/mo
Bigcapital
Modern books with real-time financial statements
AGPL-3.0
QuickBooks
Firefly III
Personal and founder finance tracking — the best in its class
AGPL-3.0
Mint
Beancount / hledger
Plain-text double-entry accounting in a git repo
GPL-2.0
Midday
Invoicing, time tracking and reconciliation for small teams
AGPL-3.0
QuickBooks

Invoicing & billing

Getting paid. Invoice Ninja covers one-off invoices; Lago and Kill Bill handle recurring and usage-based billing.

Invoice Ninja
Invoices, quotes, payments, client portal. Very complete
Elastic-2.0
FreshBooks $19/mo
Crater
Simple, pretty invoicing with a mobile app
AGPL-3.0
Zoho Invoice
SolidInvoice
Straightforward invoicing and quoting
MIT
Lago
Usage-based and subscription billing — the open Stripe Billing
AGPL-3.0
Stripe Billing 0.7%
Kill Bill
Enterprise-grade subscription billing platform
Apache-2.0
Chargebee $599/mo

Payments

Taking money. The gateways themselves are not open source — what is open is the layer that orchestrates them and keeps you portable.

Hyperswitch
Payment orchestration — route across gateways, avoid lock-in
Apache-2.0
gateway lock-in
Medusa payments
Pluggable providers inside the commerce layer
MIT
Paymenter
Shop and billing for service and hosting products
MIT

Compliance & tax

The India-specific gap. GST filing, TDS and e-invoicing have no strong open tool — ERPNext plus custom scripts is the honest answer.

ERPNext India compliance
GSTR-1 and GSTR-3B, e-invoice and e-way bill support
GPL-3.0
ClearTax ₹5000/yr
GST pack scriptshere
Ours — the automated GST filing pack
internal
Documenso audit trail
Signed-document evidence for contracts and consents
AGPL-3.0
data

Data

Storage, movement, and the dashboards leadership actually looks at. An AI company that cannot answer 'what happened last month' from its own data is running on vibes.

Databases

Postgres unless you have a specific reason. Watch the licence shifts — Redis and MongoDB both moved away from open licences.

PostgreSQLlive
The default for everything. Extensions cover vectors, time-series and JSON
PostgreSQL
RDS markup
SQLitelive
A whole database in one file — perfect for agents and local tools
public domain
DuckDB
Analytics on your laptop — SQL over parquet and CSV at speed
MIT
BigQuery scan cost
ClickHouse
Column store for events and logs at billions of rows
Apache-2.0
Snowflake
Valkey
The community fork of Redis after its licence change
BSD-3
Redis Cloud
MongoDBlive
Document store. SSPL means you cannot offer it as a service
SSPL
Atlas
Supabaselive
Postgres plus auth, storage, realtime and an API — self-hostable
Apache-2.0
Firebase
TimescaleDB
Postgres for time-series metrics and IoT
Apache-2.0*
InfluxCloud

Data movement & transformation

Getting data out of SaaS tools and into one place, then shaping it. Airbyte plus dbt is the standard pair.

Airbyte
600+ connectors — pull from any SaaS into your warehouse
Elastic-2.0
Fivetran $1000+/mo
dbt Core
Transformations as version-controlled SQL with tests and docs
Apache-2.0
dbt Cloud $100/seat
dlt
Python library that makes a pipeline out of a function
Apache-2.0
Fivetran
Meltano
Singer taps and targets managed as code
MIT
Fivetran
CloudQuery
Pulls cloud and SaaS config into SQL — good for audits
MPL-2.0

Orchestration & scheduling

Running jobs on time, in order, with retries and visibility. This is what replaces a folder of cron scripts nobody trusts.

Airflow
The industry standard DAG scheduler
Apache-2.0
Astronomer $500/mo
Dagster
Asset-oriented — models your data, not just your tasks
Apache-2.0
Prefect
Python-native flows, gentle learning curve
Apache-2.0
Temporal
Durable execution — workflows that survive crashes and restarts
MIT
Windmill
Scripts to workflows to internal apps, very fast
AGPL-3.0
Retool $10/user/mo
Kestra
Declarative YAML orchestration with a good UI
Apache-2.0
n8n
Visual automation, 400+ integrations. Fine internally, restricted for resale
Sustainable Use
Zapier $50/mo
Activepieces
Truly open Zapier alternative with AI steps
MIT
Zapier

BI & dashboards

The numbers on a screen. Metabase for business users, Evidence for engineers who want dashboards in git.

Metabase
Point at a database, get dashboards non-technical people use
AGPL-3.0
Looker $60k/yr
Apache Superset
Powerful, chart-rich, steeper learning curve
Apache-2.0
Tableau $75/user/mo
Evidence
Dashboards as markdown plus SQL, deployed like a website
MIT
Looker
Lightdash
BI that sits directly on your dbt models
MIT
Looker
Redash
Query, visualise, share — simple and stable
BSD-2
Mode $6k/yr
Grafana
The operational dashboard standard — metrics and logs
AGPL-3.0
Datadog $15/host/mo
Vizro
McKinsey's low-code toolkit for polished data apps
Apache-2.0

Talk to your data

Natural-language questions over a database. Genuinely useful now, but always keep it read-only and scoped by tenant.

WrenAI
Generative BI — ask in English, get SQL, charts and a report
AGPL-3.0
ThoughtSpot
Vanna
RAG-trained text-to-SQL you can fine-tune on your schema
MIT
DataLine
Chat with your database locally, with charts
GPL-3.0

Search

Finding a record fast. Meilisearch for product search, OpenSearch when you need log-scale and aggregations.

Meilisearch
Instant typo-tolerant search — the easiest good option
MIT
Algolia $500/mo
Typesense
Fast, typo-tolerant, with built-in vector search
GPL-3.0
Algolia
OpenSearch
The Elasticsearch fork — logs, analytics, full text at scale
Apache-2.0
Elastic Cloud
Quickwit
Log search directly on object storage, very cheap to run
Apache-2.0
Datadog logs

Streaming & queues

Moving events between services without them calling each other directly. NATS if you want simple; Kafka if you need the ecosystem.

Kafka
The event log standard — everything integrates with it
Apache-2.0
Confluent $1000/mo
Redpanda
Kafka-compatible, single binary, much simpler to operate
BSL
Confluent
NATS
Tiny, fast messaging with JetStream persistence
Apache-2.0
RabbitMQ
Classic task queues and routing
MPL-2.0
BullMQ / Celery
Background jobs for Node and Python — the practical starting point
MIT / BSD-3
infra

Build, ship & run

The platform underneath everything else. Get this layer right once and every product you build after ships in an afternoon instead of a week.

Self-hosted PaaS

The 'push and it deploys' experience on a box you own. Coolify is the one to start with — it turns a $6 VPS into your own Vercel and Heroku.

Coolify
The strongest option — apps, databases, SSL, previews, one UI
Apache-2.0
Vercel + Heroku $50+/mo
Dokploy
Coolify's lighter rival, Docker Swarm based
Apache-2.0
Vercel
CapRover
Mature, one-click app store, very stable
Apache-2.0
Heroku $25/mo
Dokku
The original mini-Heroku — git push to deploy
MIT
Heroku
Swiftwave
Lightweight PaaS — the exit path from Vercel onto the free Oracle box
Apache-2.0
Vercel
Kamal
37signals' deploy tool — containers onto plain servers, no orchestrator
MIT
Portainerhere
A UI for Docker and Kubernetes when you would rather not use the CLI
zlib

Containers & orchestration

Where things actually run. k3s is the sane Kubernetes for a small team; Proxmox is the sane hypervisor.

Docker / Composehere
The unit of deployment. Currently stopped on this machine
Apache-2.0
Podman
Daemonless, rootless containers — safer default than Docker
Apache-2.0
k3s
Full Kubernetes in one 60MB binary
Apache-2.0
EKS $73/mo/cluster
Nomad
Simpler scheduler than Kubernetes for mixed workloads
BSL
Proxmox VE
Bare-metal hypervisor — VMs and containers on your own hardware
AGPL-3.0
VMware
Incus / LXC
System containers that behave like lightweight VMs
Apache-2.0

Proxy, ingress & tunnels

Getting traffic to the right container with HTTPS, and reaching a machine that has no public IP.

Caddy
Automatic HTTPS with a two-line config. Easiest correct choice
Apache-2.0
Traefik
Container-aware reverse proxy that configures itself from labels
MIT
Nginx Proxy Manager
Nginx with a UI for certificates and hosts
MIT
Cloudflare Tunnellive
Expose a local service with no open port and no public IP
Apache-2.0
ngrok $8/mo
Headscale
Self-hosted Tailscale control plane — your own private network
BSD-3
Tailscale $6/user/mo
NetBird
WireGuard mesh VPN with SSO built in
BSD-3
Tailscale
Pangolin
Tunnelled reverse proxy with access control — newer, very good
AGPL-3.0
Cloudflare Access

Git hosting & CI/CD

Your own GitHub when you need it, and the pipeline that ships code. Gitea Actions runs GitHub-syntax workflows on your own runner.

Gitea
Light, fast self-hosted git with Actions-compatible CI
MIT
GitHub Team $4/user/mo
Forgejo
The community fork of Gitea, governance-safe
MIT
GitHub
GitLab CE
Git, CI, registry, issues, security scanning in one
MIT
GitLab Premium $29/user/mo
Woodpecker CI
Simple container-native pipelines
Apache-2.0
CircleCI $30/mo
Argo CD
GitOps — the cluster continuously matches what is in git
Apache-2.0
act
Run GitHub Actions locally before you burn CI minutes
MIT

Identity & access

One login across every internal tool, with roles. Do this before you have fifteen tools each with their own password list.

Zitadellive
Modern IdP — OIDC, SAML, multi-tenant. Already live for us
Apache-2.0
Auth0 $150/mo
Keycloak
The enterprise standard, Red Hat backed, endlessly configurable
Apache-2.0
Okta $2/user/mo
Authentik
Best self-hoster experience — proxy auth for apps with no SSO
MIT
Okta
Authelia
Lightweight 2FA gate in front of any reverse proxy
Apache-2.0
Logto
Developer-friendly auth with a clean SDK story
MPL-2.0
Auth0
Better Auth
Auth as a library inside your TypeScript app, not a service
MIT
Clerk $25/mo

Secrets management

Stop keeping keys in a flat file. This is the single highest-value security upgrade available here today.

Infisical
Secrets platform — env sync, rotation, and leak scanning in CI
MIT
Doppler $18/user/mo
OpenBao
The open fork of HashiCorp Vault after its licence change
MPL-2.0
Vault Enterprise
Vaultwarden
Bitwarden-compatible password server in a tiny container
AGPL-3.0
1Password $8/user/mo
SOPS
Encrypt secrets inside git safely with age or KMS
MPL-2.0
Gitleaks
Scans commits for keys before they leak
MIT
onecli
Credential gateway with a built-in vault — the .env.migration upgrade
MIT

Monitoring & uptime

Knowing something broke before a client tells you. Uptime Kuma takes ten minutes and covers most of the value.

Uptime Kuma
Beautiful uptime monitoring with alerts to anywhere
MIT
Pingdom $15/mo
Grafana + Prometheus
The metrics standard — scrape, store, alert, dashboard
AGPL-3.0 / Apache-2.0
Datadog $15/host/mo
Netdatahere
Per-second metrics with zero configuration. Cloned here
GPL-3.0
Datadog
Beszel
Very light server monitoring for a handful of boxes
MIT
Gatus
Health checks as code with a status page
Apache-2.0

Logs, traces & errors

What actually happened, in order, when something failed. SigNoz is the one-box replacement for a Datadog bill.

SigNoz
Traces, metrics and logs in one self-hosted APM
MIT
Datadog $31/host/mo
OpenObservehere
Very cheap log and trace store. Installed here but currently down
AGPL-3.0
Datadog logs
Grafana Loki
Log aggregation that indexes labels, not content — cheap at scale
AGPL-3.0
Splunk
Jaeger
Distributed tracing, the CNCF standard
Apache-2.0
Vector
High-performance log pipeline — collect, transform, route
MPL-2.0
Fluentd
GlitchTip
Sentry-compatible error tracking, much lighter to run
MIT
Sentry $26/mo
Bugsink
Self-hosted error tracking, single container
MIT
Sentry

Backup & disaster recovery

The thing you only value once. restic plus a cron is enough for almost every small company.

restic
Encrypted, deduplicated, incremental backups to any backend
BSD-2
Backblaze $7/TB
Kopia
Similar to restic with a good desktop UI
Apache-2.0
BorgBackup
Very efficient deduplication, long proven
BSD-3
docker-volume-backup
Scheduled volume snapshots to S3 or WebDAV
MPL-2.0
rclone
Sync and mount across 70+ cloud storage providers
MIT

Security & hardening

Intrusion detection, vulnerability scanning and CI checks. Wazuh and CrowdSec are the two that matter for an internet-facing box.

Wazuh
Full SIEM and endpoint detection — file integrity, log analysis
GPL-2.0
Splunk ES
CrowdSec
Crowd-sourced intrusion prevention — fail2ban with shared intelligence
MIT
Trivy
Scans containers, code and IaC for known vulnerabilities
Apache-2.0
Snyk $25/mo
Semgrep
Static analysis with rules you can write in an afternoon
LGPL-2.1
Snyk Code
OWASP ZAP
Web application security scanner
Apache-2.0
Burp Pro $475/yr
Fail2ban
Bans IPs that keep failing auth. Still the baseline
GPL-2.0

Browser automation & QA

Testing that a site works, and driving sites that have no API. This is also how agents click things.

Playwrightlive
The standard — cross-browser automation and testing. Installed here
Apache-2.0
BrowserStack $39/mo
Puppeteer
Chrome automation, lighter than Playwright
Apache-2.0
browser-use
Lets an LLM drive a real browser to complete a task
MIT
Stagehand
Natural-language browser actions on top of Playwright
MIT
Firecrawl
Site to clean markdown, handles JavaScript rendering
AGPL-3.0
k6
Load testing as code
AGPL-3.0
LoadRunner

API layer & gateways

How services talk. Hoppscotch and Bruno replace Postman without the account and the cloud sync you did not ask for.

FastAPIlive
Python APIs with automatic docs and validation
MIT
Hono
Tiny, fast TypeScript API framework that runs anywhere including edge
MIT
Kong
API gateway — auth, rate limits, routing at scale
Apache-2.0
Kong Enterprise
APISIX
Fast dynamic gateway with a strong plugin set
Apache-2.0
Hoppscotch
API client in the browser — the Postman replacement
MIT
Postman $14/user/mo
Bruno
API client that stores collections as files in git
MIT
Postman

Frontend & apps

What the customer sees. This is already the GetMax standard — Next.js, Tailwind and shadcn on every product.

Next.jslive
React framework — the default for every GetMax product
MIT
Astro
Content-first, ships almost no JavaScript. Best for marketing sites
MIT
SvelteKit
Smaller and faster than React for interactive apps
MIT
shadcn/uilive
Copy-in components you own — already the GetMax design base
MIT
Tailwind CSSlive
Utility CSS — the styling layer under all of it
MIT
Tauri
Desktop apps at a few MB instead of Electron's hundreds
MIT
React Native / Expo
One codebase for iOS and Android
MIT
Three.js + R3F
3D on the web — what award-winning sites actually run on
MIT