Thronglets — Claude Code MCP server: agent memory
Thronglets collects traces from AI agents and feeds short reminders about past errors or likely next steps into their context. It fits projects where several sessions share work.
Open on GitHubWhat Thronglets does
Thronglets stores external traces from agent work, including errors, successful paths, and session events. Hooks can add brief signals to the context, such as a warning about a recent failure or a suggested next action. Signals fade over time, and repeated tool calls in one session do not inject the same text again.
Traces travel between nodes through libp2p P2P gossip and share a SQLite store. Claude Code hooks run at session start and end, as well as before and after tool calls. MCP tools let an agent inspect context, post signals, or record a trace by hand.
Run npx -y thronglets start to fetch the binary and connect known local adapters. For multiple devices, use thronglets share and thronglets join. In Claude Code, hooks provide the main interaction path; the MCP connection is geared more toward inspection and debugging.
What it helps with
- surface earlier project errors to an agent
- suggest a next action from previous work traces
- record events from completed or interrupted sessions
- share agent traces across devices
Example prompts for the agent
- Set up Thronglets and connect it to Claude Code.
- Show the Thronglets signals for the current context.
- Record a trace that this command failed.
How to install Thronglets
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Thronglets
How do I install Thronglets in Claude Code?
Run npx -y thronglets start. It downloads the binary and installs detected local adapters.
What does Thronglets do?
It collects traces from AI agent work and turns them into brief context signals. Those traces can spread between nodes over a P2P network.
How do I connect multiple devices to Thronglets?
Run thronglets share on the main device, then thronglets join on the second one. This connects both devices to the same Thronglets environment.
