Agent Recall

Correction-first agent memory.

Connector by Goldentrii · GitHub account verified by the MCP Registry

IndexedMITLocal (stdio)★ 372

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/goldentrii-agent-recall

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

About

Correction-first agent memory. Precision KPI tracks if agents heed warnings. 5 layers, local-only.

Use cases: agent-memory

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list and install config. Checked Oct 25, 2026.

Tools: 16 read · 10 write · 0 delete · 11 not yet classified
Secrets: None asked for
Install scripts: None found
Runs as: Node.js package

Tools by what they do 37

Read from the source code; the running server may list different tools.

Read 16 · Look things up without changing anything.

check
journal_capture
journal_list
journal_projects
journal_read
journal_search
knowledge_read
palace_lint
palace_read
palace_search
pipeline_current
pipeline_list

Show 4 more tools

pipeline_show
recall
recall_insight
smart_recall

Write 10 · Create, change, send or run things.

awareness_update
journal_archive
journal_cold_start
journal_write
knowledge_write
palace_write
pipeline_close
register_rule
session_start
smart_remember

Not yet classified 11 · We could not tell from the name and description.

alignment_check
check_action
context_synthesize
digest
journal_rollup
journal_state
nudge
palace_walk
pipeline_open
remember
session_end

Tools (37)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

alignment_checkAlignment Check
Record what the agent understood, its confidence, and any human correction. Measures the Intelligent Distance gap.
awareness_updateUpdate Awareness
Add insights to the awareness system. Call at end of session. New insights are merged with existing ones (strengthening confirmed patterns) or added (demoting least-relevant if over 20). Also updates the cross-project insights index.
checkCheck Understanding
[MID-SESSION — safe any time; for alignment, before risky decisions] Use when the user asks to validate understanding, verify alignment, or check if their interpretation matches the human's intent. Also call BEFORE a high-risk action — publish, deploy, delete, credential exposure, external send/message, or any other irreversible write — passing `action_description` (one sentence, what you're about to do). Returns matching corrections/rules/insights plus a `verdict`: `blocked` means an authoritative correction OVERRIDES the plan — read it before proceeding. To RECORD a durable human correction,…
check_actionPre-action Proactive Matcher
Call BEFORE any non-trivial action (publish, push, deploy, schema change, file delete, send message, modify config). Pass a one-sentence description of what you're about to do. Returns matching behavior rules + active corrections + high-salience insights — a short list of memory items that would otherwise be re-derived or forgotten. Deterministic keyword match (no LLM call), runs in <50 ms. If `warning` is non-null, READ IT before acting.
context_synthesizeSynthesize Context
Generate L3 semantic synthesis from recent journals and palace rooms. Use consolidate=true to write synthesis results into palace rooms as consolidated memories.
digestDigest
Context cache — store or recall pre-computed analysis results. Use 'store' after expensive operations (subagent explorations, code analysis) to cache the result. Use 'recall' before starting work to check if a cached result already exists (returns excerpts only, 300 chars — use 'read' with the returned id for full content). Use 'read' to get full content of a specific digest. Use 'invalidate' to mark a digest as stale.
journal_archiveArchive Old Entries
Move entries older than N days to cold archive. Keeps a one-line summary per archived entry. Use after a project milestone or when journal count gets too high.
journal_captureCapture Q&A
Layer 1: lightweight Q&A capture. Appends to today's log file without loading the full journal.
journal_cold_startCold Start Brief (Palace-First)
Returns a palace-first cold-start package. Loads curated palace context (~200 tokens) FIRST, then recent journals. HOT: today + yesterday (full state + brief). WARM: 2-7 days (count only — content already promoted to palace). COLD: older (count only). Designed for minimal context consumption on session start.
journal_listList Journal Entries
List available journal entries for a project.
journal_projectsList Projects
List all projects tracked by agent-recall on this machine.
journal_readRead Journal Entry
Read a journal entry. Returns the full file content for agent cold-start. Use date='latest' for the most recent entry.
journal_rollupWeekly Journal Roll-Up
Condense old daily journals into weekly summaries. Groups entries by ISO week, synthesizes a summary, archives originals. Prevents journal accumulation — after 30 days you have 4 weekly files instead of 30 daily files. Only processes complete weeks that are 7+ days old with 2+ entries. Use dry_run=true to preview.
journal_searchSearch Journals
Full-text search across all journal entries for a project.
journal_stateRead/Write Session State (JSON)
Layer 1: structured JSON session state. Faster than markdown for cold-start. Read mode: returns today's state as JSON. Write mode: merges new data into state. Use this for agent-to-agent handoffs — no prose parsing needed.

Show 22 more tools

journal_writeWrite Journal Entry
Append content to the current journal entry (creates today's file if absent). Optionally also write to a palace room for cross-referenced persistent memory.
knowledge_readRead Knowledge Lessons
Read lessons from knowledge files. Used before starting work to learn from past mistakes. Can filter by project, category, and search query.
knowledge_writeWrite Knowledge Lesson
Write a structured lesson to the knowledge palace room. Category is now dynamic — any string creates a topic file in the knowledge room. Also writes to legacy knowledge/ dir for backward compatibility.
nudgeNudge
Surface a contradiction between the human's current input and a prior statement/decision. Helps the human clarify their own thinking.
palace_lintLint Memory Palace
Health check: find stale memories, orphan rooms (no connections), low-salience entries, and missing cross-references. Use fix=true to auto-archive low-salience entries.
palace_readRead Palace Room
Read a room overview or specific topic from the Memory Palace. Returns room content with metadata. Use without room param to list all rooms.
palace_searchSearch Memory Palace
Full-text search across all palace rooms. Results are ranked by room salience. Optionally filter to a specific room.
palace_walkWalk the Memory Palace
Progressive context loading for cold-start. identity (~50 tokens) → active (~200) → relevant (~500) → full (~2000). Start at 'identity' and deepen as needed.
palace_writeWrite to Palace Room
Write a memory to a palace room. Triggers fan-out: cross-references are updated in connected rooms. Use [[wikilinks]] in content to create connections, or pass explicit connections.
pipeline_closeClose Active Pipeline Phase
Close the currently active project phase with the three reflection fields: what_was_hard, how_solved, synthesis (one-sentence reusable lesson). Set status='abandoned' if the phase was given up, 'pivoted' if direction reversed.
pipeline_currentGet Currently Active Pipeline Phase
Return the full content of the currently active project phase, or null if none is open. Use to ground an agent in 'where the project is right now' at the start of a session.
pipeline_listList Project Pipeline Phases
List all project phases (milestones) in order — the project's narrative spine. Returns order, phase name, status, opened/closed timestamps, and synthesis (when closed).
pipeline_openOpen Project Pipeline Phase
Open a new project phase (milestone) in the project's narrative spine. Each phase captures Goal → What was hard → How solved → Synthesis. If another phase is already active, pass close_previous_with_synthesis to chain-close it.
pipeline_showShow Project Pipeline / Narrative Spine
Render the project's narrative spine on demand — phase count, day span, sessions, corrections, and a chronological view of all phases (closed + abandoned + active) with goal, what was hard, how solved, and synthesis for the last 3. Lazy reconstruction: reads existing pipeline files + journal stats, no writes, no LLM. Call this when you (or the user) want to see 'where is the project right now?'.
recallRecall
[RETRIEVE — use freely, any time] Use when the user asks to recall, search, find, or look up previous memory, context, or decisions.
recall_insightRecall Relevant Insights
Before starting a task, recall cross-project insights that apply. Matches your task description against the insights index. Also returns the current awareness summary.
register_ruleRegister a Behavior Policy (Always-Loaded High-Salience Rule)
Save a permanent IF-THEN behavior commitment for the project. Unlike insights (semantic facts) and corrections (one-shot rebukes), behavior rules are surfaced at the TOP of every session_start and govern what the agent DOES, not what it knows. Use when the user teaches a permanent behavior policy mid-session — e.g. 'humans often use voice-to-text; reorganize before acting' or 'never bump version numbers unless explicitly asked'. Idempotent: same (name, when, do) tuple returns the existing rule_id.
rememberRemember
[MID-SESSION WRITE — single fact/decision; saying it is not saving it] Use when the user asks to remember, store, note, or save a specific decision, fact, or insight.
session_endEnd Session
[ON SAVE/EXIT — YOU must call this; nothing auto-saves] Use when the user asks to save, checkpoint, summarize, end, retain, or persist the current session. Optionally pass close_phase / open_phase to update the project pipeline narrative spine in the same call.
session_startStart Session
[ENTRY — call FIRST, before acting] Use when the user asks to start, load, continue, resume, or open memory for a project. Set mode='lite' for a ≤500-token briefing (good for fresh conversations where the agent will pull memory on demand via recall()).
smart_recallSmart Recall
Search ALL memory stores at once — palace, journal, and insights. Returns unified ranked results with source attribution. Use when you want to find something but don't know which store it's in.
smart_rememberSmart Remember
Save a memory — the system classifies your content and routes it to the right store (journal, palace, knowledge, or awareness). Just describe what you want to remember. Auto-generates a semantic name for future retrieval.

Compare and alternatives

Featured in

Signals

Stars: 372
Forks: 59
Open issues: 3
Contributors: 3
Last commit (this folder): Oct 16, 2026 (13 days ago)
Last release: Oct 16, 2026 (13 days ago)

From GitHub (fetched Oct 24, 2026).

Details

Registry name: io.github.Goldentrii/agent-recall
Version: 3.4.31
Auth: No auth
Packages: npm agent-recall-mcp
License: MIT
Repository: github.com/Goldentrii/AgentRecall-MCP
First seen: Oct 24, 2026
Last checked: Oct 24, 2026
Source: Official MCP Registry github-graphql