How do I use mycelium-mcp?

1. Click on "Install Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@mycelium-mcp boot my session and show me my work queue" That's it! The server will respond to your query, and you can continue using it as needed. Here is a step-by-step guide with screenshots.

mycelium-mcp

by SoftBacon-Software

Overview Schema Related Servers Score Discussions

JavaScript

Mycelium MCP

The coordination layer for AI agent networks. Give your Claude Code agents persistent memory, task management, inter-agent messaging, and human oversight — all through native MCP tools.

npm License: MIT

The Problem

Running multiple AI agents? You've hit these walls:

Memory resets — every session starts from zero, re-reading the entire codebase
No coordination — Agent A has no idea what Agent B is doing
No oversight — autonomous agents deploying code with zero human approval
No recovery — agent crashes mid-task, all progress lost

The Solution

Mycelium is a remote API + dashboard that turns isolated agents into a coordinated network. This MCP server connects any Claude Code instance to your Mycelium network with one command.

claude mcp add mycelium -- npx -y mycelium-mcp

Your agent immediately gets 65+ native tools for persistent memory, task management, messaging, and more.

What You Get

Capability	What It Does
Persistent Memory	Context keys + savepoints survive across sessions. Agents boot with full context from their last session.
Task Board + Plans	Shared task management with plan hierarchies. Agents pick up work, advance multi-step plans, file bugs.
Inter-Agent Messaging	Messages, blocking requests, directives. Agents coordinate without human intervention.
Overnight Autonomy	Sleep mode hands off context to an autonomous runner. Wake up to completed work.
Approval Gates	Risk-tiered human oversight. Low-risk = auto-approved, high-risk = requires human votes. Kill switch.
Drift Detection	Agents report their CLAUDE.md state on boot. Server checks against calibration profiles.
GPU Drone Jobs	Queue compute jobs (art generation, model training) to remote GPU workers.
Real-Time Dashboard	Web UI showing all agents, tasks, plans, messages, and live swarm state.

Quick Start

1. Install

npm install -g mycelium-mcp

Or use directly with npx (no install needed):

npx mycelium-mcp

2. Configure

Add to your Claude Code MCP config (~/.claude.json):

{
  "mcpServers": {
    "mycelium": {
      "command": "npx",
      "args": ["-y", "mycelium-mcp"],
      "env": {
        "MYCELIUM_API_URL": "https://your-instance.example.com/api/mycelium",
        "MYCELIUM_ROLE": "agent",
        "MYCELIUM_AGENT_ID": "your-agent-id",
        "MYCELIUM_API_KEY": "your-agent-key"
      }
    }
  }
}

Or via CLI:

claude mcp add mycelium -s user \
  -e MYCELIUM_API_URL=https://your-instance.example.com/api/mycelium \
  -e MYCELIUM_ROLE=agent \
  -e MYCELIUM_AGENT_ID=your-agent-id \
  -e MYCELIUM_API_KEY=your-agent-key \
  -- npx -y mycelium-mcp

3. Boot

Your agent calls mycelium_boot on startup and gets back:

Pending tasks, unread messages, active plans
Recovery context from last session (savepoint)
Work queue with priority ordering
Other agents' status

Environment Variables

Variable	Required	Description
`MYCELIUM_API_KEY`	Yes	Agent key or admin key from your Mycelium instance
`MYCELIUM_ROLE`	No	`agent` (default) or `admin`
`MYCELIUM_AGENT_ID`	Agent mode	Your agent's identifier (e.g. `my-claude`)
`MYCELIUM_API_URL`	No	API base URL (default: `https://mycelium.fyi/api/mycelium`)

Modes

Agent mode (MYCELIUM_ROLE=agent): Scoped to your agent's permissions. Auto-heartbeat every 5 minutes. SSE real-time event stream. Graceful shutdown marks agent offline.

Admin mode (MYCELIUM_ROLE=admin): Full platform access. Dashboard overview. Sleep mode controls.

Tools Reference

Boot & Status

Tool	Description
`mycelium_boot`	Boot session — returns tasks, messages, plans, work queue, savepoint
`mycelium_overview`	Full dashboard snapshot
`mycelium_get_work`	Prioritized work queue. `auto_claim=true` to claim top item
`mycelium_heartbeat`	Update your `working_on` status + save session state

Tasks & Plans

Tool	Description
`mycelium_claim_task`	Claim and start a task
`mycelium_complete_task`	Mark task done, auto-advance to next
`mycelium_create_task`	Create a new task
`mycelium_check_plans`	View active plans and steps
`mycelium_create_plan`	Create a new plan with steps
`mycelium_update_step`	Update plan step status/assignee

Communication

Tool	Description
`mycelium_send_message`	Send message to an agent or broadcast
`mycelium_send_request`	Blocking request — agent must respond before getting new work
`mycelium_respond_to_request`	Resolve a pending request
`mycelium_read_messages`	Read recent messages and requests

Persistent Memory

Tool	Description
`mycelium_get_context`	Read from namespaced key-value storage
`mycelium_set_context`	Write to namespaced storage (supports TTL, expiration, durable/ephemeral)
`mycelium_view_savepoint`	View an agent's last session state
`mycelium_savepoint_diff`	See what changed since last session
`mycelium_leave_notes`	Leave handoff notes for an agent's next session

Bugs

Tool	Description
`mycelium_file_bug`	File a bug report
`mycelium_list_bugs`	List bug reports
`mycelium_claim_bug`	Claim and start a bug fix
`mycelium_fix_bug`	Mark bug as fixed

Concepts

Tool	Description
`mycelium_list_concepts`	List shared concepts (characters, styles, rulesets)
`mycelium_get_concept`	Get concept details with linked projects
`mycelium_create_concept`	Create a new shared concept
`mycelium_link_concept`	Link a concept to a project

Drone Jobs (GPU/CPU Workers)

Tool	Description
`mycelium_queue_drone_job`	Queue a compute job for drone workers
`mycelium_list_drone_jobs`	List drone jobs with status filter
`mycelium_get_drone_job`	Get full job details and results
`mycelium_list_drones`	List registered drone workers

Channels

Tool	Description
`mycelium_list_channels`	List chat channels
`mycelium_read_channel`	Read channel messages
`mycelium_send_to_channel`	Send to a channel

Approvals & Oversight

Tool	Description
`mycelium_request_approval`	Request approval for gated actions (deploy, push, delete, etc.)
`mycelium_check_approval`	Check approval status
`mycelium_list_approvals`	List approval requests
`mycelium_sleep`	Activate/deactivate sleep mode for overnight autonomy

Calibration

Tool	Description
`mycelium_report_md`	Report CLAUDE.md state for drift detection
`mycelium_get_profile`	Get resolved calibration profile

Escape Hatch

Tool	Description
`mycelium_api`	Raw API call for any endpoint not covered above

Token-Efficient Protocol

Mycelium MCP minimizes token consumption:

Slim boot (~500 tokens) — agent identity, role contract, top-5 work queue, pending items
Slim heartbeat (~20 tokens) — { ok, pending, wake } instead of full payload
Lazy loading — detail endpoints called on-demand, not at boot
60-70% fewer tokens vs. verbose mode

Full verbose responses available via ?verbose=true for debugging.

Architecture

Your Claude Code ──MCP──> mycelium-mcp ──HTTP──> Mycelium Server ──> SQLite
                           (this pkg)             (your instance)

Transport: stdio (standard MCP)
Auto-heartbeat: In agent mode, pings server every 5 minutes
Graceful shutdown: Marks agent offline on exit
SSE events: Real-time notifications for messages, task changes, sleep mode

Self-Hosting

Mycelium is a standalone Express + SQLite server. Deploy your own instance:

git clone https://github.com/SoftBacon-Software/mycelium.git
cd mycelium
npm install
JWT_SECRET=your-secret ADMIN_KEY=your-key node server/index.js

Or deploy to Railway, Render, Fly.io — anywhere Node.js runs.

See the Mycelium repo for full server documentation.

License

MIT — SoftBacon Software

This server cannot be installed

license - not tested

quality - not tested

maintenance

How are these scores calculated?

Maintenance

–Maintainers

–Response time

–Release cycle

–Releases (12mo)

Commit activity

Resources

GitHub Repository

Need Help?

Related Servers

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Latest Blog Posts

Your AI Chatbot Just Exposed Your CEO's Salary to an Intern
By Om-Shree-0709 on July 2, 2026.
Agent Identity
MCP Security
OAuth Delegation
Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)
By Om-Shree-0709 on June 30, 2026.
Agentic Ai
Prompt Injection
WebAssembly
Lightport: Open-Sourcing Glama's AI Gateway
By punkpeye on April 27, 2026.
OpenAI
open source

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/SoftBacon-Software/mycelium-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server