Honcho Plugin for OpenCode
Long-term memory for OpenCode: preferences and project decisions survive restarts, context wipes and fresh chats.
- 62 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- In cloud mode session text, commands and file edits go to Honcho; secrets in command arguments are redacted, but conversation content is not filtered.
Install
- Works in
Manual install
opencode plugin add "@honcho-ai/opencode-honcho"The OpenCode 2.x command from step 2 of the README. For 1.x use `opencode plugin "@honcho-ai/opencode-honcho" --global`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- In cloud mode session text, commands and file edits go to Honcho; secrets in command arguments are redacted, but conversation content is not filtered.
This is third-party code. Review the repository files before installing.
What it does
The plugin builds Honcho memory into OpenCode. Hooks inject stored context into each prompt, and significant agent activity (shell commands, file edits, delegated tasks) is recorded into the session. Command arguments that may carry credentials are redacted, keeping only the executable name. The agent gets the `honcho_search`, `honcho_chat` and `honcho_create_conclusion` tools, and a `honcho-memory` skill is copied into the OpenCode skills directory. It tells the agent when to pull memory before work and when to save a durable conclusion. OpenCode projects map to Honcho workspaces, and sessions can be scoped to a directory, repository, Git branch or shared globally.
Who it is for. OpenCode users tired of re-explaining stack, conventions and past decisions to the agent in every new chat.
Good fit when
- A project runs for weeks and the agent must recall why SQLite was chosen over Postgres.
- The same preferences (small pull requests, testing style) keep getting repeated after every restart.
- Memory has to follow the Git branch: the `git-branch` strategy separates context of different tasks.
- Accumulated local OpenCode session history should be loaded into Honcho with `/honcho:import`.
Not a fit when
- Conversations with the agent must not leave the machine: in cloud mode, import and capture send text to Honcho.
- Tasks are one-off and there is nothing to carry between sessions.
- Another agent is in use: the plugin is built only for OpenCode 1.x and 2.x.
Example request
Recall which session strategy we picked for this repository and save a conclusion on why we dropped global memory.Limitations
Cloud mode needs a Honcho API key and an account at app.honcho.dev. The key and settings live in `~/.honcho/config.json`. A self-hosted Honcho is reached through `baseUrl`, but inside Docker `localhost` may not point at your machine. Older installs keep the directional observation mode until you choose, and switching modes does not migrate earlier conclusions: use `/honcho:import` for that. A confirmed import sends conversation text from local sessions to Honcho.
How to disable. Remove the `@honcho-ai/opencode-honcho` entry from the `plugins` (2.x) or `plugin` (1.x) key in the OpenCode config. Delete the `honcho-memory` skill folder and `~/.honcho/config.json` separately.
FAQ
How do I install the Honcho plugin in OpenCode?
On OpenCode 2.x run `opencode plugin add "@honcho-ai/opencode-honcho"`; on 1.x add the `--global` flag to `opencode plugin`. Then run `/honcho:setup` and enter your API key.
Where are the settings stored?
In `~/.honcho/config.json`. OpenCode values sit under `hosts.opencode`, while the shared API key and server address live at the file root.
How do unified and directional modes differ?
Unified keeps conclusions in the user shared collection, visible to other agents too. Directional keeps a separate view of the user for the OpenCode agent. New installs get unified.
Can it work without the cloud?
Yes. Put the address of your own Honcho deployment into `baseUrl`, for example `http://127.0.0.1:8000`. A key is needed only if that server requires authentication.
Which commands appear in OpenCode?
`/honcho:setup` configures the connection, `/honcho:status` shows state, `/honcho:settings` and `/honcho:config` display and change settings, `/honcho:import` loads session history.
Discussion
No questions or reviews yet.
Related
PluginsHoncho plugins from Plastic Labs: Claude Code memory that survives sessions and skills for wiring the Honcho SDK into your app.
Honcho memory plugin for OpenClaw: separate models of the user and the agent and honcho tools beside the built-in memory.
An OpenCode plugin that keeps project decisions and agreements across context compactions and sessions with no extra model calls.
An OpenCode plugin that runs a team of parallel agents with separate sessions, a shared task board and messaging
Orchestrator plugin for OpenCode and Codex CLI: multiple agents, hooks, built-in MCPs, ultrawork mode and hash-anchored edits
Honcho MCP gives an agent memory about a user: it stores conclusions from conversations and answers questions about them.
README in short
Installation is a single plugin command, then `/honcho:setup` configures the cloud or a local server. The `recallMode` setting decides whether the plugin injects context itself or the agent reaches memory through tools.