OpenSpec MCP Server: Spec-Driven Development Skill

    1

    Your AI agent writes a plan you can approve before it writes any code.

    Free

    1 useSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    OpenSpec MCP Server: Spec-Driven Development Skill

    Example session with this skill installed

    Use OpenSpec to propose a new user authentication feature called user-auth. Start by creating the change and showing me the tasks.

    • Read your context and instructions
    • Compiled the openspec mcp server

    I have initialized the new change user-auth for the user authentication feature. I also generated the initial specification files and retrieved the task list for your review.

    Feature Proposal: user-auth

    FileStatusDescription
    proposal.mdCreatedHigh-level summary of the authentication flow.
    design.mdCreatedTechnical architecture and data structures for auth.
    tasks.mdCreatedStep-by-step implementation plan.

    Proposed Task List (tasks.md)

    1. Define User schema with email and hashed password fields.
    2. Implement /api/auth/register endpoint with validation.
    3. Implement /api/auth/login endpoint with JWT issuance.
    4. Create middleware to protect private routes using the JWT.
    5. Add unit tests for the authentication controller.

    Next steps

    • Review the design.md file in the openspec/changes/user-auth/ directory.
    • Run openspec_validate once you have customized the requirements.
    • Approve the tasks to begin implementation.
    Proposed Task List (tasks.md

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Stop AI from forgetting what you agreed or building features you never reviewed. With this skill, when you ask your agent to add a feature, it first writes a clear plan (proposal, design and task list), checks the plan for problems, and shows it to you. Only after you approve does it build, task by task, and it can tell you how far along each change is. When you confirm the work is done, it archives the change.

    You can ask things like "use OpenSpec to propose and validate dark mode", "what's the OpenSpec status of this project?" or "list our open changes".

    Included: step-by-step instructions, setup for Claude Desktop, Codex and IBM Bob, four worked examples and a tool cheat sheet. Needs Node.js (npx) and an MCP-capable agent.

    Source and MCP server: https://github.com/igor-olikh/openspec-mcp-server — by Igor Olikh

    Requires Node.js/npx. File cache may require manual refresh via openspec_refresh_cache if files are edited outside of the MCP tools. Archive operations modify files without recovery.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Requires

    OpenSpec MCP
    @fission-ai/openspec
    Node.js
    @igor-olikh/openspec-mcp-server

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access

    File Scopes

    openspec/**

    MCP tools from the openspec server run shell via npx; read/write files only in the project openspec/ folder; network to the npm registry to download @igor-olikh/openspec-mcp-server and @fission-ai/openspec. No credentials needed.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    This skill integrates the OpenSpec CLI directly into your agent's workflow, forcing it to validate designs and create structured task lists before writing code. This prevents the common 'hallucination' where agents dive into implementation without considering edge cases or existing project architecture.

    The skill is built on the Model Context Protocol (MCP) and is compatible with any MCP-enabled client, including Claude Desktop, IBM Bob, and Codex. It requires a local Node.js environment with `npx` capability to execute the OpenSpec commands.

    Yes, by running the `openspec_init` tool, you can establish a structured spec directory in any existing codebase. The agent can then use the `openspec_archive` tool to merge new feature specs into your existing project documentation.

    You will receive the MCP server configuration details and the necessary tool definitions to connect your agent to the OpenSpec CLI. This includes access to tools for initializing projects, managing change folders, validating schemas, and archiving completed tasks.

    While the skill automates spec management, it relies on a local file cache for performance. If you manually edit files in the `openspec/` directory outside of the agent's view, you should trigger the `openspec_refresh_cache` tool to ensure the agent is working with the most current data.

    No. OpenSpec is a CLI and file-based workflow.

    The README documents Claude Desktop, IBM Bob and Codex. Other MCP clients using the standard mcpServers config may work but are untested.

    The skill does not. The MCP server runs locally over STDIO and runs the OpenSpec CLI via npx.

    New in Workflow Automation