Atlas Pipeline
Run a structured seven-stage pipeline—from intent through critique and optimize—inside MCP so agents decompose work, explore variants, and refine plans before you ship code.
Overview
Atlas Pipeline is a MCP server for the Build phase that runs a seven-stage workflow from Intent through Optimize so agents plan, variant, critique, and refine work with Git context.
What is this MCP server?
- Seven explicit stages: Intent, Context, Git, Decompose, Variants, Critique, Optimize
- npm package atlas-pipeline-mcp (v1.0.2) with stdio transport
- Optional OLLAMA_BASE_URL plus OpenAI or Anthropic API keys for model backends
- Git-aware stage for tying pipeline output to repository context
- Critique and Optimize stages for iterative refinement before coding
- Seven named pipeline stages: Intent, Context, Git, Decompose, Variants, Critique, Optimize
- Server version 1.0.2 on npm identifier atlas-pipeline-mcp
- Three configurable model backends via env: Ollama, OpenAI, Anthropic
What problem does it solve?
One-shot agent prompts leave you with vague specs, no variants, and no critique loop when you are solo PM and developer.
Who is it for?
Indie builders who want a repeatable planning and refinement ritual inside MCP before large feature pushes.
Skip if: Simple CRUD tweaks that need a single file edit with no decomposition or variant exploration.
What do I get? / Deliverables
After connecting Ollama or cloud keys, the pipeline returns decomposed, critiqued, and optimized plans you can execute in Git-backed implementation steps.
- Structured task decomposition and variant sets from agent runs
- Critiqued and optimized implementation plans tied to repo context
- Repeatable seven-stage workflow callable from MCP clients
Recommended MCP Servers
Journey fit
The canonical shelf is Build with a PM tilt because the pipeline orchestrates how work is scoped, decomposed, and improved during implementation planning. PM subphase matches Intent→Context→Decompose→Variants flows that produce execution-ready breakdowns rather than a single API hook.
How it compares
Structured multi-stage planning MCP, not a deployment or monitoring integration.
Common Questions / FAQ
Who is Atlas Pipeline for?
Solo builders and small teams using MCP agents who want a formal Intent-to-Optimize pipeline with Git-aware planning.
When should I use Atlas Pipeline?
Use it when a feature is ambiguous, needs multiple solution variants, or deserves critique before you commit to implementation.
How do I add Atlas Pipeline to my agent?
Install atlas-pipeline-mcp from npm with stdio transport and configure OLLAMA_BASE_URL and/or OPENAI_API_KEY and ANTHROPIC_API_KEY as needed.