Free & open source·Self-host with your own API keys·Apache 2.0 licensed
Your knowledge graph · 23 sources
0 in scope
0 sources⌘↵
The pipeline
Uploads, recordings, images, and repository imports flow through one ingestion pipeline — OCR, transcription, chunking, embeddings — into Postgres + pgvector and a knowledge graph. Every feature that answers a question queries the same index, with citations.
You took the meeting. You saved the PDF. You wrote the Notion doc. But when you need to answer “what did our users actually care about?” — it’s buried. Launchstack fixes the last mile.
Without Launchstack
×Open 6 tabs just to answer one question
×Re-listen to calls because search can’t find audio
×Gmail threads forgotten after two weeks
×ChatGPT hallucinates — it doesn’t know your context
With Launchstack
✓One place for files, recordings, transcripts, and notes
✓Transcription + search across every mp3 and mp4
✓Import files, ZIP archives, and GitHub repos in one drop
✓Cited answers — click the chip, see the exact paragraph
Everything plugs in
If it has words, sound, or pixels, Launchstack reads it.
Drag in files of almost any format — transcription, OCR, and chunking are handled automatically. App connectors are on the roadmap; today everything comes in through upload and import.
Gmail
Threads, labels, attachments. Scope by label.
Planned
Notion
Pages, databases, nested blocks.
Planned
Google Drive
Docs, Sheets, PDFs — folder-scoped.
Planned
Slack
Channels + DMs. Pick what’s in scope.
Planned
GitHub
Issues, PRs, READMEs, your own gists.
Planned
Dropbox
Files + Paper docs, folder-scoped.
Planned
Linear
Issues, projects, cycles — searchable.
Planned
Calendar
Events + attendees give calls context.
Planned
How it works
Three steps. Zero setup.
Drop in, check what matters, ask. The animation above is an illustration of that flow.
01
Drop everything in
Drag in files, ZIP archives, call recordings, or a GitHub repo URL. Launchstack handles OCR, transcription, and indexing.
02
Check what matters
Your sources appear in a clean rail. Tick the ones you want in scope for this question. Leave the rest alone.
03
Ask. Get cited answers.
Type your question in plain English. Launchstack synthesizes across every checked source — and every claim links to the exact paragraph or timestamp.
What ships today
Built for founders and developers.
One ingestion pipeline
Upload → transactional outbox → worker → OCR / transcription → chunking → embeddings. Every document takes the same durable path into Postgres + pgvector.
Cited answers
Ask across the sources you select. Retrieval combines vector and keyword search, and every claim in the answer links back to the passage it came from.
Runs on your infrastructure
A Docker Compose stack: Next.js app, worker, Postgres + pgvector, S3-compatible storage, and compute services for transcription, document conversion, and DOCX editing.
Pricing
Open source. Nothing is for sale today.
Launchstack is free, Apache 2.0–licensed software. Self-hosting is the offering: run the whole stack on your own infrastructure with your own API keys. A managed cloud version is on the roadmap — there is no paid plan and no billing yet.
AVAILABLE NOW
Self-hosted
$0 + your keys
Fork the repo and run your own stack.
Full source under Apache 2.0
Bring your own API keys — you pay providers directly
Yes — the whole repository is Apache 2.0–licensed on GitHub: the Next.js app, the worker, the TypeScript retrieval engine, and the compute services (transcription, document conversion, DOCX editing). Clone it and run the full stack yourself with Docker Compose.
Do you train on my data?
Launchstack does not train models on your data. Retrieval and synthesis call the LLM providers you configure with your own API keys — those are the only calls that leave a self-hosted deployment.
Where does my data live?
Self-hosted, everything stays in your own infrastructure: documents and derived data (chunks, embeddings, graph rows) in your Postgres + pgvector database, raw files in your S3-compatible object storage. Deleting a document removes its database rows; the files in your object storage remain under your control.
Do I need my own API keys?
Yes — self-hosted Launchstack runs on your own Google AI key, or any endpoint speaking the OpenAI chat-completions protocol. You are not locked into our pricing and you maintain full control over your AI usage and costs.
Are the app connectors (Gmail, Notion, Slack…) available?
Not yet — connectors are on the roadmap and none are implemented today. What ships now is manual upload and import: files, ZIP archives, audio/video recordings, and GitHub repository import.
Built in the open
The engine is open source. Read it. Fork it. Run it.
The whole system lives on GitHub under Apache 2.0 — the Next.js app, the worker that runs the pipeline, the TypeScript retrieval engine, and the compute services. Audit the prompts, swap the model, self-host the whole thing.