MCP Server¶

MultiSpec includes a Model Context Protocol (MCP) server for integration with AI coding assistants.

Overview¶

The MCP server provides a standardized interface for AI assistants like Claude Code and Kiro CLI to interact with MultiSpec projects.

Installation¶

The MCP server is built as a separate binary:

# Build
make build

# The binary is at:
bin/multispec-mcp

Or install directly:

go install github.com/plexusone/multispec/cmd/mcp-server@latest

Configuration¶

Claude Code¶

Add to your Claude Code MCP configuration:

{
  "mcpServers": {
    "multispec": {
      "command": "multispec-mcp",
      "args": []
    }
  }
}

Kiro CLI¶

Add to your Kiro steering configuration:

mcp:
  servers:
    - name: multispec
      command: multispec-mcp

Transport¶

The MCP server uses stdio transport by default, communicating via stdin/stdout with JSON-RPC 2.0 messages.

Working Directory¶

The MCP server operates relative to the current working directory. It looks for:

docs/specs/ directory containing MultiSpec projects
Individual project directories with multispec.yaml

Available Tools¶

See MCP Tools for the complete list of available tools.

Project Tools¶

Tool	Description
`list_projects`	List all projects
`get_project_status`	Get project readiness

Spec Tools¶

Tool	Description
`get_spec`	Get spec content
`get_eval`	Get evaluation results
`run_eval`	Run evaluation against rubric
`synthesize`	Generate specs from sources
`reconcile`	Generate unified execution spec
`approve`	Approve a spec
`export`	Export to target system

Draft Authoring Tools¶

Tool	Description
`start_draft`	Initialize a new draft
`get_draft`	Get current draft content
`update_draft`	Save draft content
`eval_draft`	Evaluate draft against rubric
`finalize_draft`	Promote draft to final spec
`discard_draft`	Delete a draft
`list_drafts`	List all drafts in a project

Example Usage¶

Once configured, you can interact with MultiSpec through your AI assistant:

User: What multispec projects are available?
AI: [calls list_projects tool]
    Found 2 projects: user-onboarding, payment-flow

User: What's the status of user-onboarding?
AI: [calls get_project_status tool]
    Project user-onboarding is NOT READY.
    - Required specs present: PASS
    - Evaluations passing: PASS
    - Approvals obtained: FAIL (pending: prd, trd)
    - Execution spec generated: FAIL