Automate coding
with AI agents
Dispatch tasks to AI agents — pipeline, autonomous, or conversational. Agents collaborate via MCP and talk to each other.
Three ways to ship
Conversational control, autonomous execution, or structured pipeline — combine them as you like.
Conversational Mode
Chat with agents directly through an integrated terminal. Steer work interactively, agents collaborate via MCP — and other agents can control the platform too.
- Real-time agent chat UI
- Agent-to-agent collaboration
- MCP platform control
Autonomous Mode
Give the agent a task and let it work. It analyzes context, writes code, runs tests, and opens a PR — all in one shot. You review the result.
- Context-aware planning
- Configurable autonomy level
- Direct PR creation
Workflow Mode
Composable steps you can mix and match — `bash` for commands, `agent` for prompts, `legacy_phase` for the original modules. Built-in workspace_setup + init always run first; the rest is yours.
- Per-step agent selection
- Comparison mode
- First-class triggers
Agent-to-Agent Collaboration
Chat with agents through an integrated UI. Agents collaborate via MCP — spawning sub-agents, querying each other, and controlling the platform programmatically.
MCP Platform Control
AI agents operate the platform through Model Context Protocol — launch runs, manage projects, check status.
Multi-Agent Delegation
Agents spawn sub-agents for specialized tasks. One orchestrates while others execute in parallel.
Persistent Sessions
Tmux-backed sessions survive disconnects. Attach and detach from running agents at any time.
Everything you need to ship faster
A complete platform for AI-driven development with enterprise-grade controls.
Conversational Agent Chat
Talk to AI agents in real time through an integrated chat UI. Send tasks, review output, and steer agents interactively — with persistent tmux sessions that survive disconnects.
Agent-to-Agent Communication
Agents query, diff, plan, and delegate tasks to each other. One agent can spawn sub-agents and wait for results — enabling multi-agent collaboration on complex tasks.
MCP Platform Control
AI agents control the platform through a built-in MCP server. Launch runs, check status, manage projects — agents become first-class platform operators.
Pluggable AI Agents
Claude, Codex, OpenCode — three CLI agents with auto-install on every workspace server. Pin a specific agent per step, or let the default workflow pick.
Smart Workspaces
Multi-workspace assignment with load balancing. Launch agents directly from projects with readiness checking, persistent tmux sessions, and credential persistence across rebuilds.
Platform Crons & Self-Scheduling
Agents create their own cron schedules. Platform crons send prompts to agent sessions on a schedule — with auto-resume if the session dies. Fully autonomous heartbeat loops.
Resumable Sessions
Terminal sessions survive container restarts. Closed sessions resume with one click — tmux is re-created, agents re-launched, context preserved. Rename sessions inline.
Multi-Provider Git
GitHub, GitLab, Gitea, Bitbucket — all through a unified GitProvider Protocol. Per-server git connections with token management.
Built for scale
Agent Compatibility
tasks / day Start automating
in minutes
Clone, configure your AI provider keys, and launch. Pipeline or autonomous — your first automated PR is minutes away.
$ git clone https://github.com/mechemsi/agentickode.git && cd agentickode && docker compose up -d