Honcho
Your memory across AI tools
Honcho by Plastic Labs gives agents a long-term memory of you that other AI tools you connect can share. Agents save what they learn, and Honcho reasons over it to build a model of your preferences, goals and ways of working. Agents ask it questions before they start. OpenBot agents keep their own memory too; Honcho is the memory that follows you to other tools. Messages that agents save go to your Honcho account.
What an agent can do with Honcho
Ask what Honcho knows about you
The chat tool answers a question about one peer from what Honcho learned across that peer's conversations. The workspace_chat tool asks across all peers. Both take a reasoning_level from minimal to max; the default is low. Use minimal for fast lookups and high or max when depth matters.
Record a conversation
The create_session tool makes a session for a conversation, and create_peer makes a peer: any participant, human or AI. The add_peers_to_session tool adds peers and sets which ones Honcho observes. The add_messages_to_session tool records both sides of each exchange.
Read stored memory quickly
The search, get_representation, get_peer_context and list_conclusions tools read what Honcho already has, with no new reasoning. They are faster than chat when the agent needs only the stored facts.
Keep the memory in order
The set_peer_card tool sets the short biographical card of a peer by hand. The schedule_dream tool starts a memory consolidation in the background. The list_workspaces and create_workspace tools find or make the workspace that holds the memory.
How to connect Honcho to OpenBot
Open Honcho in OpenBot
Press Open in OpenBot on this page, or find Honcho in the app's Apps tab. Get the app first if it is not on this computer.
Sign in
Press Connect and finish the sign-in in your browser. OpenBot keeps the token on the computer that runs OpenBot. You can disconnect at any time.
Add your API key
Press Connect and enter this value. OpenBot keeps it on the computer that runs OpenBot.
- API key - Create a key at app.honcho.dev.
Give an agent a task
Ask an agent for work that needs Honcho, such as one of the example prompts below. The agent uses the plugin's tools to do it.
Example prompts3
HonchoAsk Honcho what it knows about how I like to work before we start.
Apps1
Honcho
Save and query long-term memory in your Honcho workspaces over one MCP server.
Information6
- Developer
- Plastic Labs
- Category
- Productivity
- Version
- 1.0.0
- Website
- honcho.dev
- Privacy policy
- app.honcho.dev/privacy
- Terms
- app.honcho.dev/tos
Questions
Does the Honcho plugin need an API key?
No. Select Connect and sign in with your Honcho account in the browser. To use a key, select Use an API key instead, and type a key from app.honcho.dev. Hosted keys start with hch-. OpenBot sends the key in the Authorization header and keeps it in encrypted storage.
Which workspace does the agent use?
Each Honcho tool that works on a workspace takes a workspace_id. The agent names the workspace on each call, and it can use list_workspaces to find it. To keep one workspace, tell the agent its name, for example in its instructions.
How is Honcho different from the memory of an OpenBot agent?
Each OpenBot agent has its own memory, which stays on the computer that runs OpenBot. Honcho is a memory that other AI tools you connect to the same account can share. Your OpenBot workspaces, conversations and files stay on the computer that runs OpenBot. Only the messages and questions that an agent sends to Honcho tools go to Honcho.
Can I run Honcho on my own computer?
Yes, but not from this listing, which connects to the hosted server. Honcho can run locally with Docker. Its MCP server is the mcp package in the Honcho repository, pointed at your own instance with HONCHO_API_URL. Over HTTP it uses port 3000 and serves MCP at /mcp. Add it in the MCP servers settings.
Sources
Every statement about Honcho comes from the pages below, checked on . They can change after that date.