Your knowledge is scattered across hundreds of files. Cloud AI won't read them privately. And it never shows you where the answer came from.
Open source · Self-hosted · Privacy-first
Your Private AI
Knowledge Companion
Ask questions about your documents. See exactly where every answer comes from. Nothing ever leaves your machine.
Core — FreeProVault — Enterprise
100%
Private
30+
File Types
1-cmd
Setup
Apache-2.0
Open Source
See it in action
90-second walkthrough: verified answers, subjects visualization, automatic memory, live sources, and TrustScore transparency.
No sign-up. No cloud. Everything stays on your machine.
How it works
One command to start. Ask a question. Get a verified answer.
1
Import your files
Drop a folder or upload files. Cerid parses PDFs, docs, code, emails, and archives automatically.
2
Ask anything
Natural language queries search all your knowledge. Adaptive retrieval matches the right strategy.
3
See which documents answered your question
Inline verification badges on every claim. Click any footnote to see the source and reasoning.
4
Gets smarter over time
Facts, decisions, and preferences are remembered. Each conversation makes the system smarter.
Connect every source of knowledge
22 source connectors out of the box. Webhook receiver, browser extension, voice notes, and first-class Apple Mail + Reminders. Everything streams into the same verified pipeline.
RSS & Atom
URL watch
Bookmarks
Browser extension
Voice notes
Clipboard
Apple Mail
Reminders
Slack / Discord
GitHub / Linear
Sentry / Stripe
Readwise / Pocket
Gmail / Outlook
Custom webhook
+ extensible
Webhooks with HMAC
Stream into Cerid from any service that fires a webhook. Constant-time signature compare, per-source tokens, rotation built in.
Browser extension
Save the page you're reading — and the passage you highlighted — to your private KB in one click. Chrome + Firefox.
Voice notes
Record a quick thought, drop it in, and it transcribes + indexes alongside everything else. ⌘⇧V opens the overlay.
Built for you
Whatever your field, Cerid adapts to how you work.
For people who need to cite sources
See which documents back up every answer — with page-level citations you can verify in seconds.
- Researchers, journalists, and lawyers who can't afford unsourced claims
- Inline footnotes link every statement to the original passage
- Cross-reference findings across thousands of files instantly
For teams that can't risk data leaving
Healthcare, legal, and finance teams get AI answers without uploading a single file to the cloud.
- Everything runs on your machine — documents never leave your network
- No telemetry by default, no cloud dependency
- Fernet-encrypted credentials and full audit logging
For builders who want full control
Self-host with Docker. 60 MCP tools + any external MCP server. 12 specialist agents + a custom-agents builder. 22 source connectors. Bring your own model or run Ollama free.
- Full REST API with versioned /sdk/v1/ surface + Python SDK
- Governed external MCP client — allowlist + per-call audit + kill switch
- Apache-2.0 licensed — fork it, extend it, ship it
Everything you need
Tap any card to learn more. Powerful enough for regulated-deployment intelligence. Simple enough to clone, run, and use.
Know where answers come from
Every answer includes inline citations so you can see exactly which document — and which passage — backs up each claim.
Streaming claim verification with inline source attribution. 4-type claim detection (factual, recency, evasion, citation). Click any footnote to see source, confidence, and reasoning. Expert mode uses frontier models for re-verification.
See verification flow→
Finds the right part of your documents
Ask a question and Cerid pinpoints the most relevant sections across all your files — even when the answer spans multiple documents.
Three RAG modes (Manual, Smart, Custom Smart) with per-chunk retrieval profiles. Hybrid BM25 + vector search. Cross-encoder reranking. Adaptive pipeline adjusts search depth based on question complexity.
See pipeline details→
Works with the AI you choose
Use Claude, GPT, Gemini, Llama, or run a completely free local model with Ollama. Switch anytime — no lock-in.
Smart capability-based routing across OpenRouter providers. Guided Ollama install wizard. 6 of 8 pipeline tasks run locally at $0. Proactive model switching on ignorance detection.
See model routing→
Gets smarter the more you use it
Cerid remembers your preferences, past decisions, and key facts — so follow-up questions get better answers without repeating yourself.
6-type memory salience scoring: empirical facts, decisions, preferences, project context, temporal events, conversational insights. Memory recall auto-injects alongside KB context.
Your data never leaves your device
Everything runs on your machine. No cloud uploads, no telemetry by default. Only the query context you choose is sent to the AI model.
Self-hosted Docker stack with ChromaDB, Neo4j, Redis — all on your machine. Optional Fernet encryption at rest. Open source Apache-2.0.
Works with files you already have
Point at a folder and Cerid indexes everything — PDFs, Word docs, spreadsheets, code, emails, even zip archives. Over 30 file types supported.
Bulk folder scan with estimation preview. Archive extraction (zip/tar). Automatic junk filtering. SSE progress streaming with pause/resume controls. 30+ file types supported.
Connect anything that produces knowledge
22 source connectors out of the box — RSS feeds, web pages, bookmarks, browser extension, voice notes, Apple Mail, Reminders, Slack, GitHub, and more.
22 source kinds across 9 families: web (RSS, URL-watch, bookmarks, browser extension), capture (voice notes, clipboard), Apple ecosystem (Mail, Reminders), chat (Slack, Discord, Teams, Matrix), dev events (GitHub, Linear, Sentry, Stripe), reader apps (Readwise, Pocket, Instapaper, Raindrop), email (Gmail, Outlook). Webhook receiver with HMAC + token auth; OAuth scaffold for Google + Microsoft. Per-source retention + quality floors.
See every source→Privacy by design
Your knowledge base stays on your machine. Telemetry is opt-in and disabled by default.
- All data stored locally — nothing leaves your machine
- Encrypted at rest for sensitive knowledge bases
- Free local AI option — no API costs required
- Open source — audit every line of code
- Smart model routing — always the right model for the task
Stored locally
Your documents, embeddings, and metadata stay on your machine.
Encrypted at rest
API keys and credentials protected with Fernet encryption.
Open source
Apache-2.0 licensed. Audit every line of code.
Start with one command
Clone the repo, run the setup script, and your private AI companion is ready.
Terminal
$ git clone https://github.com/Cerid-AI/cerid-ai.git
$ cd cerid-ai && ./scripts/start-cerid.sh
$ cd cerid-ai && ./scripts/start-cerid.sh