Recall
Fully local project memory for Claude Code: a session log and a condensed summary with no API keys and nothing sent out
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Reads session transcripts that may contain code and secrets
- Writes the log and summary to project files
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/recallDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Run one by one in the Claude Code chat
/plugin marketplace add raiyanyahya/recall
/plugin install recall@recallInstall the plugin: /plugin marketplace add raiyanyahya/recall, then /plugin install recall@recall. Save memory with /recall:save.
Other ways from the author
/plugin marketplace add raiyanyahya/recall
/plugin install recall@recallNo pip install: the summarizer is vendored and stdlib-only. OpenCode setup is separate.
This is third-party code. Review the repository files before installing.
What it does
Recall solves the cold-start problem: Claude Code begins every session fresh, while Recall keeps a local log of the work and condenses it into a resume-ready summary right on your machine. Two files go into .recall/: history.md with the full session log and context.md with a short summary of the goal, what was done, which files were touched and where you stopped. The condensing is done by a vendored classical Python summarizer, with no model call, keys or local model. Nothing leaves your machine, and the memory is plain markdown you can read and share across agents.
Who it is for. For people running Claude Code on a subscription who do not want to re-explain the project every session.
Good fit when
- You are tired of re-explaining the project at the start of each session
- You want memory that spends no tokens and never goes to the network
- You need to share context between Claude Code and OpenCode on one repo
Not a fit when
- You need semantic search over a large base rather than session summaries
- The built-in CLAUDE.md and resume are enough
Example request
Save the session memory with /recall:save and at the next start bring up where we left offLimitations
The summary is made by a deterministic local algorithm rather than an LLM, so it is simpler than a model-made one. Session logs contain code, paths and sometimes secrets, but stay on your machine. Claude Code is the first-class path; OpenCode setup is separate and opt-in.
How to disable. Uninstall via /plugin uninstall recall@recall; the project's .recall/ folder can be removed manually.
Security check
- Reads session transcripts that may contain code and secrets
- Writes the log and summary to project files
README in short
The README describes Recall as fully local project memory for Claude Code that solves the cold start with no keys or external models. Into .recall/ it writes history.md as an append-only log and context.md as a condensed summary made by a vendored classical Python summarizer. The author contrasts Recall with CLAUDE.md and resume: it is an automatic, deterministic record of what each session did. Install is via its own marketplace or locally, with opt-in OpenCode support on the same markdown files, and the memory is fenced as untrusted data.
FAQ
Does Recall spend tokens?
No. The summary is built by a local Python summarizer with no model call, and a compact context.md at start saves session tokens.
Does data go to the network?
No. Logs and summaries stay in .recall/ on your machine, there is no external API, and it works offline.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything