Announcing the CoreStory + Devin MCP Playbook

Give Devin a real source of truth

We’re excited to announce the release of our CoreStory + Devin MCP usage playbook. This guide is for CoreStory users who want Cognition’s Devin to work from the same source of truth as their engineers: CoreStory’s understanding of the codebase. Through MCP, Devin can query specs, contracts, flows, and patterns directly—leading to more accurate code, stronger tests, and the ability to finish longer-running tasks.

Why this integration matters

Agents like Devin do their best work when they can see the full context, not just the current file or ticket. By connecting Devin to CoreStory via MCP, you let it cross-check what it’s about to do against the real behavior and contracts of your system. The playbook packages this into a repeatable pattern you can adopt in roughly a sprint, without overhauling your existing workflows.

What this playbook helps you do

With this playbook, Devin learns to:

• Consult CoreStory specs first, then cross-check spec vs implementation for gaps
• Work in a test-first style, generating or extending tests that are derived from actual specifications
• Handle longer (>5 ACU) tasks like rewrites and migrations with a much higher success rate

What’s inside

The playbook includes:

• Step-by-step setup instructions for wiring Devin to CoreStory via MCP
• A tooling map of CoreStory MCP tools (projects, conversations, documents) and when to use each
• A ready-to-use start prompt pattern for Devin that enforces project verification, spec consultation, and test-first work
• Workflow templates for spec/code gap analysis, parity testing, discrepancy reporting, and cost control (retrieve vs generate)

Who this is for

If you’re already using Devin in production or serious pilot projects and you want it to behave more like a senior engineer working from shared docs and specs—not a guessing engine—this playbook is for you.

How to get started

Get the playbook, follow the quick-start section to connect Devin to CoreStory in a single environment, then choose one representative workflow (ticket resolution, rewrite, or parity testing) to pilot. Use accuracy, task completion rate, and test coverage as your primary evaluation metrics.