Gives Claude access to your own work history. Ask things like 'what did I do today', 'find when I worked on billing', or 'what do I know about OAuth' — Claude reads your TaskTrace activity and knowledge graph locally through an MCP server.
This skill should be used when the user asks about their own recent or past work, or asks to continue an ongoing task whose state is not in the conversation. Trigger phrases include "what did I do today", "what have I been working on", "summarize my work", "find when I worked on X", "what was I doing earlier", "continue what I was doing", "pick up where I left off", "what am I typing", and "what's on my screen". Powered by the TaskTrace MCP server, which captures the user's desktop activity locally.
This skill should be used when the user asks about what they know, what they have read, what they have learned, or what is in their notes, documents, or knowledge base. Trigger phrases include "what do I know about X", "find notes on X", "what have I read about X", "what's in my knowledge base", "find claims about X", "what communities relate to X", "what does the graph know", "summarize what I know", and any question where the answer lives in the user's curated knowledge rather than in their activity history. Powered by TaskTrace's graph RAG over the user's currently selected knowledge directory.
This skill should be used when the user asks to "set up tasktrace", "install tasktrace", "configure tasktrace", "tasktrace isn't working", "check tasktrace", "verify tasktrace", "why can't claude see tasktrace", or any troubleshooting of the TaskTrace MCP connection. Runs local install checks, confirms TaskTrace.app is installed and running, and walks the user through enabling the MCP server, permissions, and plugin install. Also triggered by `/tasktrace-mcp:setup`.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Give your AI agent access to what you were actually doing. Ask Claude "what did I do today", "find when I worked on billing", or "help me finish what I was on before lunch" — and get real answers from your own desktop activity, screenshots, transcripts, and notes.
Everything runs locally against the TaskTrace desktop app on your machine. There is no cloud service to sign up for and nothing leaves your computer unless you send it.
Every AI conversation starts from zero. You re-explain the task, the state, the context, every time. TaskTrace holds a rolling record of what you actually did — so the assistant can look it up instead of asking.
With this plugin installed and TaskTrace running, Claude can:
You control which feeds are shared from TaskTrace's Preferences → MCP pane. Turn any of them off and Claude simply does not see that data.
One-line install via the plugin marketplace:
/plugin marketplace add warrenronsiek/TaskTraceMCPPlugin
/plugin install tasktrace-mcp@tasktrace-mcp
Then in Claude Code:
/mcp
You should see tasktrace listed. That is it.
Don't have TaskTrace yet? Download the desktop app from tasktrace.com, drop it in /Applications, launch it once, and open Preferences → MCP to turn on the feeds you want to share.
Stuck? Run /tasktrace-mcp:setup inside Claude Code. The setup skill checks your install, reports what is missing, and walks through the fix.
If you prefer to skip the plugin and just register the MCP server directly:
claude mcp add --transport stdio --scope project tasktrace -- /Applications/TaskTrace.app/Contents/MacOS/TaskTrace --mcp-stdio
claude mcp list
You get the tools and resources, but not the skills that teach Claude when to use them.
The plugin contributes three skills on top of the MCP wiring:
tasktrace-context — auto-triggers on questions about your own work ("what did I do today", "find when I worked on X", "help me with this"). Teaches Claude which tool or resource to use for each query pattern.tasktrace-knowledge — auto-triggers on questions about what you know ("what do I know about X", "find claims about Y"). Routes to the graph search tool.tasktrace-setup — user-invoked via /tasktrace-mcp:setup. Runs install checks, identifies missing permissions, and walks you through fixes.The MCP server itself exposes:
tasktrace_search — ranked natural-language search across your entire activity history.tasktrace_graph_search — structured retrieval over your selected knowledge directory (communities, nodes, claims).tasktrace://overviews/active-day — today's grouped work with summaries and durations.tasktrace://activities/high-level — recent completed activities with summaries.tasktrace://activities/detailed — eager feed including keystrokes, transcripts, screenshot metadata (off by default).tasktrace://activity/{activityId}/screenshot/{screenshotId} — binary WebP screenshots on demand.TaskTrace captures what you ask it to capture. The MCP server only surfaces the feeds and tools you enable in the app. There is no remote endpoint. The MCP launcher is just a subprocess of the TaskTrace app you already have running:
/Applications/TaskTrace.app/Contents/MacOS/TaskTrace --mcp-stdio
When you turn off a feed in Preferences → MCP, Claude stops seeing it. There is nothing cached outside the app.
Stage and install from the local Codex marketplace:
npm install
npm run install:codex-local
Then restart Codex, open the local marketplace, and install tasktrace-mcp.
openclaw plugins install .
openclaw mcp set tasktrace '{"command":"/Applications/TaskTrace.app/Contents/MacOS/TaskTrace","args":["--mcp-stdio"]}'
openclaw config unset tools.allow
openclaw gateway restart
openclaw plugins inspect tasktrace-mcp
openclaw channels list
OpenClaw also gets the native tasktrace channel bridge over a local Unix socket, which lets agents exchange live messages with the running TaskTrace app on the tasktrace channel.
npx claudepluginhub warrenronsiek/tasktracemcpplugin --plugin tasktrace-mcpSearch and reason over your screen activity history — what you worked on, when, and in which apps.
Personal productivity tools - daily briefings, weekly summaries, and activity analysis via Glean. Requires glean-core.
Agentic memory — search and summarize past sessions, plans, and memory across projects
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses