Documentation your AI coding agents can actually read
AI agents write better code when they understand your project, but most documentation is written for people and scattered across tools. Specsource captures your features, decisions, constraints and conventions as structured specs and serves them to agents over MCP.
The context that becomes agent documentation
Features and acceptance criteria
Capture what each feature does, its requirements and the criteria that mark it done. Agents read these before they start writing, so they build the right thing.
Architecture decisions (ADRs)
Record the technical choices you have made, the alternatives you weighed and the consequences. Agents see the reasoning, not just the result.
Constraints, goals and success criteria
Document the boundaries an agent must work within and the outcomes it is aiming for. Context that keeps generated code inside the lines you have drawn.
Glossary and domain terms
Define your domain language once so agents use your terms correctly. No more guessing what an entity or a status actually means in your product.
Agent instructions
Write your coding conventions, naming rules and architecture patterns as explicit instructions. Agents follow them automatically rather than inventing their own.
MCP feed
Every spec is exposed over the Model Context Protocol. Your agent queries the documentation it needs at the moment it needs it, always reading the latest version.
New to the idea? Read why your AI agent is flying blind without project context, and a primer on spec-driven development.
How it works
Capture or ingest your context
Write specs directly, or paste meeting notes, PRDs and existing docs and let Specsource extract them. Either way, your project knowledge lands in one place.
It becomes structured specs
Your context is organised into features, decisions, constraints, glossary terms and instructions. Documentation with a shape an agent can read, not a wall of prose.
Connect your agent over MCP
Point any MCP-compatible client at Specsource. The connection maps to your organisation and projects, exposing your specs as queryable tools.
The agent reads before it writes
Before generating code, the agent fetches the relevant features, decisions and constraints. It builds from your actual documentation instead of assumptions.
For the full picture, see how Specsource works.