Claudetor
Claudetor Mascot

Your agents
finally remember.

Claudetor gives your AI agents instant, scalable, and secure long-term memory. Drop it into your stack and never lose context again.

$ pip install claudetor
4.9k GitHub Stars 1.2M Downloads
CA 5nVUwabp8q5txwUsua7iM8apk22PnoHmaAfb5Gbmpump
Streamflow
WORKFLOW

From install to local
recall in four steps.

Claudetor abstracts away the complexity of vector databases and embedding models.

01

Initialize

Run `claudetor init` to generate your local config and set up the SQLite vector store.

02

Sync Context

Use `claudetor sync ./src` to chunk and embed your entire codebase automatically.

03

Retrieve

Call the API or CLI to semantic search across your memory state instantly.

04

Build Agents

Pass the retrieved context directly into your LLM prompt for intelligent execution.

import { Claudetor } from '@claudetor/sdk';

const memory = new Claudetor({
  projectId: 'agent-alpha',
  provider: 'local'
});

// Auto-retrieve context for user prompt
const context = await memory.recall(prompt);
CONTEXT

Every session starts
with full context.

Don't waste tokens summarizing previous interactions. Claudetor automatically injects relevant historical data into the context window.

  • Persistent state across restarts
  • Automatic token optimization
  • Drop-in replacement for LangChain memory
Query Latency (ms)
SPEED

Search years of
codebase in milliseconds.

We built a custom Rust-based indexing engine that runs locally. It embeds and searches through gigabytes of text without network latency.

  • < 50ms average retrieval time
  • Built-in local embedding models
  • Zero cloud dependencies required
txt
pdf
csv
INTEGRATION

One RAG API.
Every machine.

Unify your data sources. Claudetor ingests PDFs, code repositories, Slack channels, and Notion docs into a single queryable memory bank.

  • 50+ native data connectors
  • Automatic metadata extraction
  • Graph-based relationship mapping

What teams build for months.
Yours in one command.

Stop reinventing the RAG wheel. Claudetor gives you production-ready memory out of the box.

Capability DIY RAG Stack Claudetor
Vector DB Setup Takes days, requires DevOps Instant local SQLite
Chunking Logic Hardcoded & breaks on code AST-aware semantic chunking
Local Execution Complex to orchestrate Default behavior
State Management Manual JSON saving Automatic persistent sessions
Cost High cloud API bills Free & local-first

Developers stopped
starting from zero.

"Claudetor completely changed how we build AI agents. We dropped 4 different libraries and replaced it with one `claudetor init` command. It just works."

S

Sarah Jenkins

Lead AI Engineer, Nexus

"The AST-aware chunking is magic. It actually understands Python classes and functions instead of blindly splitting text. Saved us 2 months of engineering."

M

Marcus Chen

CTO, DataFlow

"Fastest local RAG setup I've ever seen. The fact that I can run the entire memory stack on my MacBook without hitting an API is a game changer for privacy."

E

Elena Rodriguez

Indie Hacker

Claudetor Logo
OPEN SOURCE

Built in the open.
Trusted by builders.

4.9k
GitHub Stars
1.2M
NPM Downloads
85+
Contributors

Free during early access.

Start building with the core open-source engine today. Cloud sync features coming soon.

COMMUNITY EDITION
$0

Everything you need to build local autonomous agents.

  • Unlimited local vector storage
  • AST-aware chunking engine
  • 50+ document loaders
  • Local embedding models
ENTERPRISE CLOUD
Coming Soon

For teams that need scalable, shared memory clusters.

  • Cloud-hosted vector DB
  • Multi-agent memory sync
  • Role-based access control
  • SOC2 Compliance

Frequently Asked Questions

Does Claudetor send my code to the cloud?
No. By default, Claudetor's vector database and embedding models run 100% locally on your machine. Your proprietary code never leaves your network unless you explicitly configure a remote LLM API (like OpenAI).
Which LLMs are supported?
Claudetor is model-agnostic. You can use it with OpenAI, Anthropic, Google Gemini, or local models via Ollama and Llama.cpp.
How is this different from LangChain?
LangChain is a broad framework for building LLM apps. Claudetor is a hyper-focused, out-of-the-box memory solution. We handle the dirty work of chunking, indexing, and retrieval automatically, so you don't have to write boilehttps://app.streamflow.finance/contract/solana/mainnet/4dhQk5etWG9iBAWHaKb8omyXFyKAStqp33HYskBefRsJ?utm_source=twitter&utm_medium=apprplate code.
Claudetor Mascot

Memory is your
agent's superpower.

Give your AI the context it needs to actually be useful. Install Claudetor today and stop starting from zero.

$ pip install claudetor