Find work and solutions from previous sessions using a staged memory search. Filter results by ID, inspect nearby context, and retrieve full observations or original tool inputs and responses only when needed.
Use this when the user asks to search memory, "did we already solve this?", "how did we do X last time?", or wants work from previous sessions.
Run score Not tested
See how to installSearches memory in stages to find work and solutions from previous sessions. It starts with search to find record IDs, then uses timeline to inspect the surrounding context. After narrowing the results, it reads detailed records with get_observations. If summaries leave questions unanswered, it uses get_tool_uses to inspect the original tool inputs and responses. The scope and level of detail depend on the information needed.
Use it to check whether you have already solved a problem or recover the steps from an earlier session. When a single record does not explain the background, it can examine nearby records as well.
Requires MCP, a system for connecting tools. If MCP is missing, the expected setup uses an installation skill to run npx claude-mem install --ide <host>. A single original tool record can contain thousands of tokens, so it is a last resort. Records from other hosts remain included unless you request otherwise.
This overview was written by AI (OpenAI Codex) from the public SKILL.md. Check the original below for the exact behavior.
AI checked the SKILL.md for operations that need care and for setup requirements. The score comes only from actually running the skill (hands-on test).
Not counted in the score. Rated by AI (OpenAI Codex) as of Oct 21, 2026. Treat it as a guide and read the original before installing.
Review the instructions the agent will receive before you install.
Layered workflow. Do not dump full observations until IDs are filtered, and do not reach for raw tool bodies until the summaries came up short.
search(query) — index with IDstimeline(anchor=ID) — nearby contextget_observations(ids=[...]) — full details for those IDs onlyget_tool_uses(ids=[...]) — the ORIGINAL tool_input / tool_response for specific tool calls. Last resort: these are unsummarized and can run to thousands of tokens each.Stamp platformSource as cursor or grok-bot on writes for this host. When reading, do not drop the other host unless asked.
If MCP is missing, run the install skill first (npx claude-mem install --ide <host>).
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Turn agent activity and transcript token counts into a cost report organized around completed outcomes. Separate estimates from measured spending, examine waste and rework, and export readable HTML with supporting JSON, CSV, and evidence.
Monitor a pull request through reviews and automated checks, verify reported issues against the code, and make focused fixes. Repeat testing and review checks until actionable comments and unresolved threads are cleared and the request is ready to merge.
Delegate a coding task to a subagent that runs build, merges the ready pull request, and continues from the updated main branch into release, connecting implementation and release work in one automated sequence.
Check a local claude-mem worker, refresh viewer-specific caches of important observations, and exclude or restore records without deleting history. Inspect rules across shared, project, and agent layers, then append conflicts to a report.
Monitor a shipped release by checking published packages, live services, installed plugins, and late CI results or user reports. Verify regressions, prepare authorized fixes, and report release health with evidence and any remaining verification gaps.
Set up claude-mem in Cursor with a local or remote worker, the process that runs its work. Follow configuration steps for remote inference through cmem.ai or a local observer using the host's login.