Stop babysitting your AI agents
You have 6 agents running across 4 machines. You're tab-switching between terminals, copy-pasting context from one agent's output into another's prompt, and restarting dead sessions at 2 AM. Every agent has its own CLI, its own auth flow, its own way of crashing silently.
Grackle is the control plane for AI coding agents. Configure once, supervise by exception.
One platform to run Claude Code, Copilot, Codex, or Goose on any environment — Docker, SSH, Codespaces, local. It handles provisioning, credentials, transport, and lifecycle. You get a CLI, web UI, and MCP server out of the box.
Grackle is pre-1.0 and still experimental. It may have unresolved security issues, annoying bugs, and broken workflows. Not recommended for use in production systems.
What makes Grackle different
Agent IPC — Parent sessions spawn children with bidirectional pipes. Structured communication between agents — no polling, no shared files, no prompt-stuffing.
Knowledge persistence — A semantic knowledge graph backed by Neo4j. One agent's architectural insight becomes another agent's context automatically. Search by concept, not keyword.
Session resilience — Environments auto-reconnect on disconnect. Suspended sessions resume where they left off. Events buffer during outages and drain on reconnect. No lost work.
Multi-vendor, one interface — Swap runtimes per persona or per task. Your orchestration doesn't break when you switch from Claude to Codex or add Copilot as a second opinion.
Plugin architecture — The server is composed of plugins that you can toggle on and off. Run the full orchestration stack or strip down to a lightweight session manager.
How it fits together
The Grackle Server is the control plane. It manages environments, sessions, tasks, and credentials. You interact with it through the chat interface, CLI, web UI, or MCP server. Inside each environment, PowerLine runs the actual agent and streams events back to the server.
Features
| Feature | Description |
|---|---|
| Chat interface | Natural language command interface — just describe what you want |
| Real-time streaming | Watch agent tool calls and output as they happen |
| Git worktree isolation | Every task gets its own branch — zero interference between agents |
| Knowledge graph | Semantic memory backed by Neo4j — agents share knowledge automatically |
| Findings | Categorized discoveries shared across agents within a workspace |
| Multi-runtime | Claude Code, Copilot, Codex, and Goose — swap freely |
| Task trees | Decompose work into parent/child subtasks up to 8 levels deep |
| Signals | SIGTERM, SIGCHLD, cascade kill, orphan adoption — kernel-style process control |
| Personas | Specialized agent configs with system prompts, tools, and model selection |
| Scheduled triggers | Cron-style automated task creation |
| Plugin system | Compose server capabilities — orchestration, scheduling, knowledge graph |
| MCP server | Expose Grackle's full API as MCP tools for any AI agent |
Scales from remote control to swarms
| Level | What you get | What you use |
|---|---|---|
| 1. Remote control | One agent, one environment, you watch it work | Sessions, environments |
| 2. Structured tasks | Break work into tasks with branches and review gates | + Workspaces, tasks, personas |
| 3. Parallel agents | Multiple agents working independently, sharing findings | + Multiple environments, findings |
| 4. Orchestrator pattern | Parent agent decomposes work and coordinates child agents via MCP | + Task trees, MCP broker, signals |
You don't need to adopt everything at once. Each level builds on the last — see the orchestration guide for details.
Next steps
- Getting Started — Install Grackle and run your first agent in 5 minutes
- Credential Setup — Configure API keys for Claude, Copilot, Codex, and Goose
- Concepts — Understand environments, sessions, tasks, and the rest of the model
- Guides — Web UI, orchestration, chat, plugins, and more