Skip to main content
Glama
agent-loop-mcp
by meharajM

Agentic Loop Memory Server ♾️

Agent Skills skills.sh

The industry-standard persistent memory and state manager for long-running agentic workflows.

Enable any AI model—especially smaller ones with limited context windows—to function with the persistence of high-end models. This project works as a two-part ecosystem: an MCP Server for state management and an Agent Skill for orchestration.

🛠 Complete Setup (Required)

For the best experience, you must install both the orchestration skill and the MCP server.

1. Install the Skill

Install the agentic-loop skill into your AI agent (Codex, Claude Code, Cursor, Gemini CLI, GitHub Copilot, and other Agent Skills hosts):

npx skills add meharajM/agent-loop-mcp@agentic-loop -g -y

Preview the skill before activation:

gh skill preview meharajM/agent-loop-mcp agentic-loop

2. Configure the MCP Server

Add the following to your `mcp_config.json`:

{
  "mcpServers": {
    "agent-loop": {
      "command": "npx",
      "args": ["-y", "@mhrj/mcp-agent-loop"]
    }
  }
}

Related MCP server: Ori Mnemos

🌟 Why this approach is unique

Unlike passive memory tools, this is an Active State Manager. It monitors word counts to trigger compaction cycles and enforces a "Self-Healing Strategy" on every failure, preventing AI agents from getting stuck in mindless loops.

📂 Project Structure

  • src/: TypeScript source for the MCP server.

  • skills/agentic-loop/SKILL.md: The instruction manual for the AI.

  • build/: JavaScript artifacts.

📄 License

ISC

Available Tools

6 tools
compact_memoryA

Empties the Active Context and appends the AI-provided summary to the Compacted History.

Inputs: session_id (required), context_summary (required).

get_tool_suggestionsB

Ask this tool if you are stuck and don't know what other tools to use.

Inputs: no named parameters.

init_loopA

Creates a new .md loop state file for the given session to start a continuous, autonomous task.

Inputs: objective (required), session_id (required).

log_stepA

Appends to the Active Context. Rejects if failed=true but no self_heal_strategy is provided. Warns if context is too large.

Inputs: action (required), failed (required), result (required), session_id (required), +1 more.

report_blockerA

Updates state to STATUS_BLOCKED and asks for human intervention when absolutely stuck.

Inputs: reason (required), session_id (required).

resume_loopC

Removes the block and adds human input context back into the loop.

Inputs: session_id (required), user_input (required).

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 6 tools

    Disambiguation5/5

    Each tool has a distinct purpose: init starts a loop, log appends steps, compact clears memory, resume unblocks, report triggers human intervention, and get_tool_suggestions offers guidance. No overlapping functionality.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern (init_loop, log_step, compact_memory, resume_loop, report_blocker, get_tool_suggestions), using snake_case throughout.

    Tool Count5/5

    With 6 tools covering the essential loop lifecycle operations, the count is well-scoped for a specialized agent-loop management server.

    Completeness5/5

    The toolset provides comprehensive coverage for managing an autonomous loop: initialization, logging, memory compaction, resumption, blocker reporting, and suggestion retrieval. No obvious gaps in the workflow.

    Maintenance

    ActivityNo data
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      B
      maintenance
      Provides persistent memory for LLM applications, enabling AI assistants to remember user preferences, facts, and conversation history across sessions.
      2
      -
    • A
      license
      Not graded
      quality
      D
      maintenance
      Provides durable memory for AI agents with structured storage, semantic search, OAuth authentication, and lifecycle controls.
      10
      Apache 2.0