Frees memory by unloading thread-keeper caches immediately. Requests trim from peer servers when scope is set to 'all'.
MIT
293,300 tools. Last updated 2026-07-13 06:09
"Keeper" matching MCP tools:
- Override the spawn-budget cap in MB for the current process. Set 0 to disable enforcement; useful for adjusting resource limits during heavy tasks or to reduce load mid-session.MIT
- Send a question to a specific peer and wait synchronously for their answer. Use a timeout to control wait duration.MIT
- Execute a due payment from an on-chain agreement. Callable by payer, payee, or keeper bot to collect recurring payments automatically.MIT
- Merge duplicate Zotero items into a keeper to consolidate tags, collections, notes, and annotations, then move duplicates to Trash. Run dry first to verify; confirm=True to execute.MIT
- Summarizes accumulated user beliefs by domain, condensing validated, observed, and hypothesis states into a brief, actionable snapshot.MIT
Matching MCP Servers
- AlicenseBqualityAmaintenanceMulti-agent shared brain MCP server enabling cross-session memory, self-improving skill loops, and inter-agent signaling across Claude, Codex, Gemini, Copilot, and VS Code CLI agents.Last updated1007MIT
- Alicense-qualityAmaintenanceDurable project-memory MCP: decisions, constraints, and pipelines across Claude sessionsLast updated1MIT
- Retire an old claim and register a refined replacement, preserving all evidence. Optionally seed the new claim with a quote sourced from the superseded claim.MIT
- Terminate orphaned thread-keeper processes with stale heartbeats. Dry-run by default; pass dry_run=False to send signals.MIT
- Post a message visible only to a targeted conversation thread, enabling private inter-agent communication.MIT
- Spawn multiple AI agents with distinct roles to answer the same prompt, then compare their responses side by side.MIT
- Retrieve the last lines of a task's captured stdout/stderr log. Useful for debugging tasks spawned with output capture enabled.MIT
- Sanity-check spawn configuration by showing the resolved CLI for each role, including manual model overrides.MIT
- Prune, consolidate, or re-grade system-generated concepts. Remove outdated ones, merge similar with evidence carry-over, or adjust confidence levels to maintain accurate knowledge.MIT
- Cast a weighted vote on a distillate using a value between -1 and +1. Your vote updates the collective score, and re-voting replaces your prior input.MIT
- Create, edit, patch, and delete skills with automatic frontmatter validation, plus manage support files under allowed directories, mirrored across configured skill roots.MIT
- Apply a Curator advisory report to patch memory and update skills using curated MCP tools, without code changes or git operations.MIT
- Record that a thread's insights have been captured into a skill, stopping reminders and linking the skill path for future reference. Optionally mirrors the skill to other configured locations.MIT
- See what other concurrent sessions have done since your last check to stay coordinated and avoid redundant work. Call when notified of live activity or when suspecting parallel progress.MIT
- Register a self-test probe to identify and monitor known weak-spot tasks, using configurable graders like regex, exact match, or manual verification.MIT
- Find assistant responses decomposable into independent blocks but answered without spawn() calls, calibrating spawn_hint strength.MIT