Task Graph Mcp
Model multi-step agent jobs with phases, prompt templates, approval gates, and coordinated sub-agents from your MCP host.
Overview
io.github.Oortonaut/task-graph-mcp is an MCP server for the Build phase that runs phased agent workflows with prompts, gates, and multi-agent coordination.
What is this MCP server?
- MCP server v0.3.0 for phased agent task workflows with prompts and explicit gates
- Multi-agent coordination so several agents share one task graph state
- Prebuilt mcpb binaries for Linux x86_64 and macOS (Intel and Apple Silicon)
- Stdio transport suitable for Claude Code, Cursor, and other local MCP hosts
- Complements mcacp when MCP triggers ACP agents inside governed phases
- Version 0.3.0 with 3 platform mcpb packages (Linux x86_64, macOS Intel, macOS aarch64)
- Stdio transport on all published packages
- GitHub Oortonaut/task-graph-mcp with SHA256-pinned release binaries
Community signal: 1 GitHub stars.
What problem does it solve?
Long agent sessions drift without phase boundaries, so you cannot enforce prompts, approvals, or handoffs between specialized agents.
Who is it for?
Solo builders running multi-step agent pipelines who need gates and phase prompts in MCP instead of ad-hoc instructions.
Skip if: Simple one-shot API MCP servers like geocoding or a single OSINT lookup with no workflow state.
What do I get? / Deliverables
You get an MCP-managed task graph with gated phases and coordinated agents so each milestone completes before the next starts.
- Phased task graphs callable as MCP tools
- Prompt and gate definitions shared across agents
- Multi-agent coordination state for long-running jobs
Recommended MCP Servers
Journey fit
Task graphs are authored while you are building agent workflows, even though the same graphs can gate Validate scope and Ship review. PM subphase fits because the server centers on phased tasks, prompts, and gates—not a single REST-style integration.
How it compares
Workflow orchestration MCP server, not a raw protocol bridge or marketplace skill pack.
Common Questions / FAQ
Who is io.github.Oortonaut/task-graph-mcp for?
Builders who delegate complex work to multiple AI agents and need phases, gates, and shared task state in MCP.
When should I use io.github.Oortonaut/task-graph-mcp?
Use it when scoping or building agent runbooks, or when you want review gates before ship steps in the same graph.
How do I add io.github.Oortonaut/task-graph-mcp to my agent?
Download the mcpb release for your OS (v0.3.0), register it as a stdio MCP server in your client config, and point tools at the graph API.