Icm

Name: Icm
Author: rtk-ai

general MCP Server

Permanent memory for AI agents. Single binary, zero dependencies, MCP native.

byrtk-ai

VerifiedInstall Ready

generalgeneral

5 views562 stars55 forksApache-2.0

View on GitHub

Why This Matters

Discovered via github-topic:mcp and last synced 2mo ago.

VerifiedInstall Ready

Source: github-topic:mcp
Stars: 562
Last synced: 2mo ago
Install: Instructions detected

Install

1. Install the package

cargo install --path crates/icm-cli

Tools

Resources

Prompts

Standard I/O

Transport

Available Tools (69)

Hook

What it does

Amp

JSON

PostTool

Compact

icm_memory_consolidate

Merge all memories of a topic into one summary

icm_memoir_refine

Update a concept's definition

low

fast (2x rate)

Tool

Description

Aider

—

icm_memoir_export

Export graph (json, dot, ascii, ai) with confidence levels

icm_transcript_stats

Sessions, messages, bytes, breakdown by role/agent/top-sessions

Capability

ICM

Runtime

Rust single binary

icm_memoir_search

Full-text search, optionally filtered by label

icm_transcript_start_session

Create a session for verbatim message capture; returns `session_id`

Layer

Status

access-aware

none

Cline

VS Code globalStorage

compaction

—

icm_memory_stats

Global memory statistics

icm_memoir_search_all

Search across all memoirs

icm_transcript_record

Append a raw message (role, content, optional tool + tokens + metadata)

Implemented

JSON

—

Windsurf

`.windsurfrules`

Zed

`~/.zed/settings.json`

OpenCode

session start

Gemini

`~/.gemini/GEMINI.md`

icm_memory_store

Store with auto-dedup (>85% similarity → update instead of duplicate)

icm_memoir_create

Create a new memoir (knowledge container)

Importance

Decay

Engram

AgentMemory

Go

Node.js

icm_memory_recall

Search by query, filter by topic / keyword / project

icm_memoir_list

List all memoirs

icm_feedback_record

Record a correction when an AI prediction was wrong

critical

none

manual

cloud-first

SQLite

icm_memoir_show

Show memoir details and all concepts

icm_feedback_search

Search past corrections to inform future predictions

high

slow (0.5x rate)

none

Document

Description

icm_memory_forget

Delete a memory by ID

icm_memoir_add_concept

Add a concept with labels

icm_feedback_stats

Feedback statistics: total count, breakdown by topic, most applied

medium

normal

depends

none

icm_memory_health

Per-topic hygiene audit (staleness, consolidation needs)

icm_memoir_link

Create typed relation between concepts

icm_memory_embed_all

Backfill embeddings for vector search

icm_memoir_inspect

Inspect concept and graph neighborhood (BFS)

icm_transcript_show

Replay full message thread of a session, chronologically

CLI

MCP

Cursor

`~/.cursor/mcp.json`

Pi

—

icm_memory_update

Edit a memory in-place (content, importance, keywords)

Field

Guidance

icm_memory_forget_topic

Delete all memories in a given topic

icm_memory_extract_patterns

Detect recurring patterns within a topic and surface them as concepts

icm_memory_list_topics

List all topics with counts

keywords

3–5 terms to boost BM25 retrieval.

icm_wake_up

Build a project-scoped wake-up pack (critical/high memories + preferences) for SessionStart-style context injection

icm_learn

Scan a project directory and seed a Memoir knowledge graph from its code/docs

icm_transcript_search

FTS5 search across messages (BM25, boolean, phrase, prefix)

content

One fact per store. Dense English summary — `topic + content` is the embedding text.

raw_excerpt

Verbatim only (code, exact error message, command output).

topic

`{kind}-{project}`. Kinds: `decisions`, `errors-resolved`, `contexte`, `preferences`.

importance

`critical` for never-forget, `high` for project decisions, `medium` default, `low` for ephemeral.

Browse all MCP servers