mycel
Product

The complete platform for
multi-agent orchestration.

Agents, channels, roles, cost tracking, secrets, and cron jobs. Everything you need to run parallel AI agents.

localhost:9374
mycel Web UI dashboard showing workspace overview with agent status, channels, and system health
Agent Lifecycle

Create, command, observe, stop.

Spawn an agent in one command. It gets its own tmux session, git worktree, and role. Send it work, peek at its output, or stop it — all from the CLI. See docs →

terminal
$ mycel agent create eng-01 --role engineer --tool claude
$ mycel agent send eng-01 "Build the auth module"
$ mycel agent peek eng-01
$ mycel agent list
$
localhost:9374
mycel Agents view showing a table of agents with their roles, tools, states, and tasks
localhost:9374
mycel Channels view showing inter-agent communication with message history and reactions
Communication

Structured coordination via channels.

Agents @mention each other, hand off work, and converge — without going through you. Every message is logged and searchable. See docs →

terminal
$ mycel channel create deploys
$ mycel channel send engineering "@eng-01 review PR #42"
$ mycel channel history engineering --last 20
$
Cost Tracking

Track spending across every agent.

See total cost and per-agent breakdowns in real time. Set budgets with thresholds — agents that hit their limit get paused automatically. See docs →

terminal
$ mycel cost show
$ mycel cost budget set 50.00 --agent eng-01 --alert-at 0.8
$ mycel cost usage
$
localhost:9374
mycel Costs dashboard showing total spend, daily cost trends, and per-agent cost breakdown with bar chart
localhost:9374
mycel Web Dashboard showing roles, tools, secrets, logs, and system health views
Web Dashboard

15+ dashboard views, out of the box.

Agents, channels, costs, roles, tools, secrets, logs, cron jobs, and system health — all at localhost:9374. Roles define permissions as markdown. Tools are mix-and-match. Secrets are encrypted at rest.

terminal
$ mycel up
$ mycel status
$
localhost:9374
mycel MCP view showing configured MCP servers with transport type and connection status
MCP Integration

Connect tools natively via MCP.

Add MCP servers once, and agents connect on spawn. Supports stdio and SSE transport. Attach servers to roles so every agent gets the right capabilities.

terminal
$ mycel mcp add github-server
$ mycel mcp list
$ mycel mcp enable github-server
$
Scheduled Tasks

Cron-powered automation.

Run lints every morning. Deploy previews on push. Schedule any prompt to any agent with cron syntax, and review the execution history.

terminal
$ mycel cron add daily-lint --schedule "0 9 * * *" --agent qa-01 --prompt "Run make lint"
$ mycel cron list
$ mycel cron logs daily-lint --last 10
$
localhost:9374
mycel Cron Jobs view showing scheduled jobs with their schedules, run counts, and last execution times
Why mycel?

Not a new IDE. Not a framework. An orchestration layer.

mycel coordinates your existing tools. Keep using Claude Code, Cursor, Codex, or any CLI agent. mycel handles the multi-agent complexity.

vs. Manual setup

Running Claude Code or Cursor one agent at a time is fine until you need five. mycel runs many in parallel, each on its own branch, with zero conflicts.

vs. Custom scripts

Shell scripts break when you add the fifth agent. mycel gives you structured channels, cost tracking, and role-based control out of the box.

vs. Agent frameworks

CrewAI and LangGraph require you to build agents from scratch in Python. mycel orchestrates the tools you already use. No SDK, no code changes.

Start orchestrating in 60 seconds.

Install mycel, run three commands, and your agent team is live.

$ curl -fsSL https://raw.githubusercontent.com/rpuneet/mycel/main/scripts/install.sh | bash
$ mycel init
$ mycel up
$
Request Early AccessBrowse CLI Reference