SkillAgentSearch skills...

mike

A tool to know what claude is doing and remember what you did when you did it.

Install / Use

claude mcp add temet-ai -- npx -y github:temet-ai/mike

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

75/100

Supported Platforms

Claude Code
Claude Desktop

Our assessment of mike

mike scores 75/100 on our quality scale, 839th of 957 AI & Machine Learning skills we index.

Its MCP Server is 8.2 KB long, well organised into 9 sections with 1 code example: a thorough specification that gives an agent plenty to work with.

It has 3 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
29/30
Structure
17/20
Description
12/15
Adoption
3/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated today, so mike is actively maintained.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 87/100, with 2 cautions from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-11-16. Automated pattern scan on 2026-11-16. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

mike compared with similar skills

All 4 of these similar skills score higher than mike; compare them before choosing.

SkillScoreStarsUpdatedFormat
mike (this skill)by temet-ai753todayMCP Server
claude-memby thedotmack100117.7ktodayCLAUDE.md
Agent-Reachby Panniantong100112.4k9d agoCLAUDE.md
Understand-Anythingby Egonex-AI10092.9k1d agoCLAUDE.md
rufloby ruvnet10081.9ktodayCLAUDE.md

Frequently asked questions

How do I install mike?
Run claude mcp add temet-ai -- npx -y github:temet-ai/mike. The install tabs above show the steps for each supported agent.
Which AI agents does mike work with?
It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
Is mike safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 87/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is mike still maintained?
The repository was last updated today, so mike is actively maintained.
<h1 align="center"><img src="docs/assets/mike-badge.png" alt="mike" width="240"></h1> <p align="center"> <a href="https://github.com/temet-ai/mike/releases"><img alt="Latest release" src="https://img.shields.io/github/v/release/temet-ai/mike?include_prereleases&amp;label=release&amp;color=eb300a&amp;labelColor=071c1e"></a> <img alt="Linux x86_64 and Apple Silicon macOS, with Linux ARM64 from the next release" src="https://img.shields.io/badge/platform-Linux%20%7C%20macOS-eb300a?labelColor=071c1e"> <a href="./LICENSE"><img alt="MIT licence" src="https://img.shields.io/badge/license-MIT-eb300a?labelColor=071c1e"></a> </p> <p align="center"> <a href="#try-it">Try it</a> · <a href="https://temet.ai/docs/mike/">User guide</a> · <a href="#how-it-works">How it works</a> · <a href="docs/user-guide/privacy.md">Privacy</a> · <a href="https://github.com/temet-ai/mike/releases">Releases</a> </p>

mike is an inspector for Claude Code, written in Rust. It records every request and answer of your sessions on your own machine, so you can open any turn in a browser, search your history, and turn a correction into something the next session starts with. It is for people who use Claude Code every day and want to see inside it.

Status: alpha, for Linux x86_64, Linux ARM64 and Apple Silicon macOS. The current release is 0.1.0-alpha.5. Expect rough edges and changes between alphas.

A captured session in mike: the system prompt, the prompt you typed, and each tool call with its result

What you can do with it

  • Read any turn as it was sent. The system prompt, the messages, the tool results and the streamed answer, grouped by session and live as they happen.
  • See what changed before it left. If a feature you turned on rewrites a request, the original and what was sent are shown side by side.
  • Ask Claude Code about its own history. mike-mcp, the companion program, gives a session tools to search earlier turns and notes.
  • Keep a correction. Tell Claude to record it as a lesson and later sessions in that project can be handed it.
  • Spot weak answers. Local pattern checks flag unbacked claims, blame-shifting and avoided work beside the response.
  • Watch and steer. Optional steerers notice a session drifting from its goal, filter rules rewrite the system prompt, and fleets run several sessions towards one goal.

A fresh install only captures: it forwards your requests unchanged and sends nothing to anyone but Anthropic. Features that rewrite a request or send text to another service stay off until you turn them on. Privacy lists what each one stores and sends.

Try it

Homebrew works on macOS and on Linux, where mike needs glibc 2.35 or newer. Install mike also covers mise and archives.

brew install temet-ai/mike/mike temet-ai/mike/mike-mcp

mike setup    # where to keep history, and which features to turn on
mike claude   # starts mike if needed, then Claude Code routed through it

At setup, keep captures in memory for a first look, or choose a PostgreSQL database to keep your history; mike can run one for you in Docker. The Watch only choice leaves every optional feature off.

Then open http://localhost:9990. Your session is under Sessions, updating as you work. Only sessions started through mike are captured. Get started walks through it.

A look around

<table> <tr> <td width="50%" valign="top"> <a href="docs/user-guide/images/request.png"><img src="docs/user-guide/images/request.png" alt="A request shown as the original and as sent, side by side, with the difference marked"></a> <br><strong>Original and sent, side by side</strong> </td> <td width="50%" valign="top"> <a href="docs/user-guide/images/flagged-response.png"><img src="docs/user-guide/images/flagged-response.png" alt="Flagged responses grouped by check and category, each linked to its session and turn"></a> <br><strong>Weak answers flagged, with the turn they came from</strong> </td> </tr> <tr> <td width="50%" valign="top"> <a href="docs/user-guide/images/sessions.png"><img src="docs/user-guide/images/sessions.png" alt="The Sessions page: search, sessions grouped by project, and the time each one was active"></a> <br><strong>Every session, searchable by project</strong> </td> <td width="50%" valign="top"> <a href="docs/user-guide/images/fleet.png"><img src="docs/user-guide/images/fleet.png" alt="A fleet graph: a writer and a reviewer session working, with the messages between them"></a> <br><strong>A fleet of sessions and the messages between them</strong> </td> </tr> </table>

Before you run it

mike has no login, so use it only on a computer nobody else uses. Its local API has no authentication: any program on the machine can read every captured prompt and answer, change mike's settings and start Claude Code sessions. Keep it on the loopback address it starts with. Remote access is not supported. SECURITY has the full trust boundary.

Your Claude login stays yours. Each person runs mike with their own Claude Code login. mike forwards the credential Claude Code sends and never stores it. Background jobs set to a claude/ model run your own claude command, so they use your subscription.

How it works

mike sits between Claude Code and api.anthropic.com. Claude Code talks to it on localhost:9988, mike forwards to Anthropic over HTTPS and keeps a copy, and the web UI on :9990 and the mike-mcp companion read that copy through a local API on :9989.

A request goes through mike to Anthropic, the original and the response are stored, and the web UI and mike-mcp read them

Named for Mike, the thinkum dinkum who ran the Moon in Heinlein's The Moon Is a Harsh Mistress. He heard every line on Luna and forgot nothing. This one settles for your Claude Code sessions.

A name clash to know about: the MkDocs tool also called mike may be on your PATH. Check with which -a mike.

More animated walkthroughs cover each part. docs/ARCHITECTURE.md has the design and its reasons.

Documentation

The user guide covers everything a user needs: install, setup, every page and command, and what to do when something looks wrong.

Reference and contributor docs. Start with the system map: one page that lists everything mike can do, where each ability lives in the code and how to verify it.

| Doc | What it holds | |---|---| | docs/SYSTEM_MAP.md | Every capability, its surface, its default and its code; the index of all other docs | | docs/FEATURES.md | Every optional feature, its default and what it does to traffic | | docs/CONFIGURATION.md | Boot config and every MIKE_* environment variable | | docs/SECURITY.md | What each bind exposes and what leaves the machine | | docs/REQUEST_FLOW.md · docs/STARTUP_FLOW.md | How a request and a boot run | | docs/SCHEMA.md | The capture database | | docs/DEVELOPMENT.md | Working from a checkout | | docs/RELEASING.md | How releases are built |

Contributing

Issues and pull requests are welcome; start with CONTRIBUTING. A source checkout runs the development stack with ./scripts/start-proxy.sh, and docs/DEVELOPMENT.md has the loop. Run scripts/test-suite fast before you send a change.

Support

If mike saved you debugging time and you'd like to say thanks: buy me a coffee.

License

MIT © Dawid du Toit

Related Skills

View on GitHub
GitHub Stars3
CategoryAI
Updated1h ago
Forks0

Languages

Rust

Trust signals

87/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

2 low