Agent Tasks
Wire your coding agent into a staged pipeline with dependencies, artifacts, and task claiming so multi-step builds do not collapse into one messy thread.
Overview
io.github.keshrath/agent-tasks is a MCP server for the Build phase that gives AI agents pipeline-style task management with stages, dependencies, artifacts, and claiming over stdio.
What is this MCP server?
- Models pipeline stages with explicit dependencies between agent tasks
- Supports artifact handoffs so each step leaves inspectable outputs for the next
- Task claiming flow to avoid two agents stomping the same work item
- Stdio npm package `agent-tasks` at registry version 1.9.9
- Fits long-running agent workflows beyond a single chat session
- Published server version 1.9.9 on npm as identifier agent-tasks
- Stdio transport; GitHub source at keshrath/agent-tasks
Community signal: 6 GitHub stars.
What problem does it solve?
Multi-step agent builds fall apart when every instruction lives in one chat with no shared task graph or artifact trail.
Who is it for?
Solo builders running repeatable agent pipelines (scaffold → implement → test → docs) who want MCP-native task state rather than copying checklists into prompts.
Skip if: Teams that only need a lightweight todo list in Notion, or builders who run single-shot prompts with no staged agent workflow.
What do I get? / Deliverables
After you register the server, your agent can drive a structured pipeline—claim tasks, respect dependencies, and pass artifacts between stages instead of re-deriving context each turn.
- MCP tools for creating and advancing pipeline tasks with dependencies
- Artifact-linked steps agents can reference across sessions
- Claim-based concurrency so parallel agents do not duplicate work
Recommended MCP Servers
Journey fit
Agent pipeline orchestration is where solo builders formalize how Claude/Codex moves from plan to shipped slices; the canonical shelf is Build because that is when MCP-backed task boards replace ad-hoc prompts. Agent-tooling is the right subphase because this server exists to manage agent work units (stages, deps, artifacts)—not generic PM docs or backend APIs.
How it compares
MCP pipeline task server for agents, not a general project-management skill or a hosted Kanban app.
Common Questions / FAQ
Who is io.github.keshrath/agent-tasks for?
It is for indie developers and agent power users who orchestrate Claude Code, Cursor, or Codex across multiple dependent build steps and need shared task and artifact state.
When should I use io.github.keshrath/agent-tasks?
Use it when a build or ship loop has clear stages (design, implement, verify) and you want the model to claim work, honor dependencies, and leave artifacts for the next stage.
How do I add io.github.keshrath/agent-tasks to my agent?
Add the npm stdio server `agent-tasks` (package version 1.9.9) to your MCP client config, restart the agent host, and verify the tools appear before running pipeline commands.