Skip to main content

​
Your AI Agent’s Long-Term Memory

Remember every conversation. Search instantly. Never repeat yourself. Bossa gives your AI agents a persistent filesystem to store and retrieve context across sessions—just like Claude Code and Manus do.
Install: pip install bossa-memory · Base URL: https://bossamemory.com · No infrastructure to run. Sign up, get an API key, start building.
Free tier: 100 MB storage, 1,000 requests/day. No credit card required. Get Started Free · Pricing · GitHub

​
The Problem: Agents Forget Everything

Every time your agent starts, it’s like Groundhog Day:
  • Re-learning user preferences every session
  • Asking the same questions over and over
  • No memory of previous work or context
  • RAG pipelines that are complex and brittle
Claude Code and Manus solved this with filesystems. Now you can too.

​
Why Filesystem Over RAG?

Leading agent platforms like Claude Code and Manus chose filesystems over traditional RAG pipelines. Here’s why:
Filesystem (Bossa)Traditional RAG
grep "user preferences"Tune embeddings, pray it works
Directory structure = organizationFlat vector space
Works like CLI tools agents knowNew abstractions to learn
Debuggable with ls, catBlack box retrieval
No embedding driftConstant re-indexing
For 80% of agent use cases, filesystems just work better. Read: Why We Chose Filesystem Over RAG →

​
Why Bossa?

​
Feels Like Local Files

bossa files ls /memory
bossa files read /user/preferences.json
bossa files grep "project alpha"
Just standard filesystem operations. No new concepts.

​
Powered by Postgres

  • Full-text search with trigrams
  • ACID transactions
  • JSON storage
  • Enterprise-ready

​
Plug-and-Play

PlatformHow to connect
Terminalexport BOSSA_API_KEY=sk-xxx then bossa files ls /
LangChainAdd Bossa MCP server to MultiServerMCPClient
Claude / CursorAdd MCP server URL + API key in headers
One line of code. That’s it.

​
Your Agent’s Personal File System

Each agent gets its own space, scoped by API key. No cross-tenant leakage.

​
How Does Bossa Help Me?

Coding Assistant

Remember project context, tech stack, and conventions across sessions.

Multi-Agent Team

Multiple agents share one workspace. Handoffs, shared state, no duplicate work.

Context Engineering

Discover what’s relevant with ls/grep, pull only what you need. Fewer tokens.
See More Examples →

​
Quick start

1

Sign up

Create an account via the CLI
2

Create workspace & key

Create a workspace and API key
3

Connect

Make your first request via MCP or REST

​
Docs MCP

Give your AI assistant (Cursor, VS Code, Claude) direct access to the Bossa documentation. Use the contextual menu at the top of any page for one-click Connect to Cursor or Connect to VS Code — or follow the manual setup instructions. Add Docs to Your IDE →

​
Documentation

Getting Started

Sign up, API key, first request.

Why Filesystem?

Filesystem vs RAG — why we chose this approach.

Dynamic Context Discovery

How Bossa fits the context-engineering pattern used by Cursor and LangChain.

Docs as MCP

Add Bossa docs to Cursor, VS Code, Claude.

CLI Reference

Auth, workspaces, keys, files.

MCP Integration

Claude, Cursor, LangChain.

REST API

Endpoint reference.

Agent Integration

Examples and patterns.

Use Cases

Coding assistant, multi-agent team, context engineering.

Self-Hosting

Run on your infrastructure.
Last modified on March 13, 2026