feat(mcp): add MCP server integration to skills and init

.claude-plugin/marketplace.json

@@ -5,7 +5,7 @@
     "url": "https://github.com/reqstool"
   },
   "metadata": {
-    "version": "0.2.0",
+    "version": "0.3.0",
     "description": "AI-assisted reqstool requirements traceability — skills and commands for managing requirements, SVCs, and filters, with optional OpenSpec integration."
   },
   "plugins": [

.github/plugin/marketplace.json

@@ -6,7 +6,7 @@
     "url": "https://github.com/reqstool"
   },
   "metadata": {
-    "version": "0.2.0",
+    "version": "0.3.0",
     "description": "AI-assisted reqstool requirements traceability — skills and commands for managing requirements, SVCs, and filters, with optional OpenSpec integration."
   },
   "plugins": [

CLAUDE.md

@@ -32,4 +32,4 @@ copilot plugin install --path ./plugins/reqstool-openspec
 
 ## Pre-commit checks
 
-Run `claude plugin validate .` before committing to catch manifest errors, missing fields, and path issues.
+See [CONTRIBUTING.md](CONTRIBUTING.md) for the full testing and contribution process.

README.md

@@ -98,7 +98,7 @@ modules:
 | `/reqstool:init` | Create or update `.reqstool-ai.yaml` configuration interactively |
 | `/reqstool:add-req` | Add a new requirement and update subproject filters |
 | `/reqstool:add-svc` | Add a new Software Verification Case and update filters |
-| `/reqstool:status` | Show requirements traceability status for system or a module |
+| `/reqstool:status` | Show requirements traceability status (uses MCP server if configured, falls back to CLI) |
 ### Skills (auto-applied)
 
 | When you're... | Skill |
@@ -129,7 +129,9 @@ Add a new Software Verification Case (SVC) to the system-level `software_verific
 
 ### `/reqstool:status`
 
-Run `reqstool status local` for the system or a specific module. Accepts an optional module name (e.g., `/reqstool:status core`); defaults to the system-level path from `.reqstool-ai.yaml`. Reqstool traverses imports and implementation config automatically.
+Show requirements traceability status for the system or a specific module. Accepts an optional module name (e.g., `/reqstool:status core`); defaults to the system-level path from `.reqstool-ai.yaml`.
+
+Uses the `reqstool` MCP server (`get_status` tool) when configured. Falls back to `reqstool status local` CLI if the server is not available, notifying you immediately when it does.
 
 ## Skill details
 
@@ -150,6 +152,25 @@ OpenSpec integration conventions, auto-applied when working with spec.md files:
 - **reqstool-openspec-conventions.md** — how to reference reqstool IDs in OpenSpec specs
 - **config-rules.yaml** — reqstool rules for openspec/config.yaml
 
+## MCP server
+
+The [reqstool MCP server](https://github.com/reqstool/reqstool-client) (reqstool ≥ 0.9.0) exposes structured tools for AI agents: `get_status`, `list_requirements`, `get_requirement_status`, and more.
+
+The `/reqstool:init` command can configure it automatically. To set it up manually, add to `.mcp.json` in your project root (project-scoped) or `~/.config/claude/mcp.json` (global):
+
+```json
+{
+  "mcpServers": {
+    "reqstool": {
+      "command": "reqstool",
+      "args": ["mcp", "local", "-p", "<your-system-path>"]
+    }
+  }
+}
+```
+
+When configured, skills like `/reqstool:status` use MCP for structured data. If the server is not available they fall back to `reqstool status local` automatically, notifying you when they do.
+
 ## Prerequisites
 
 - [reqstool CLI](https://github.com/reqstool/reqstool-client) (`pipx install reqstool`)

plugins/reqstool/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "reqstool",
   "description": "AI-assisted reqstool requirements traceability — skills and commands for managing requirements, SVCs, and filters.",
-  "version": "0.2.0",
+  "version": "0.3.0",
   "author": {
     "name": "reqstool",
     "url": "https://github.com/reqstool"

plugins/reqstool/skills/reqstool/add-req/SKILL.md

@@ -83,9 +83,10 @@ For reqstool fundamentals (imports, filters, implementations), see `reqstool-ove
    Add the new requirement ID to the `filters.<urn>.requirement_ids.includes` list in the
    relevant module's `<module.path>/requirements.yml`.
 
-6. **Verify with reqstool**
+6. **Verify the new requirement is tracked**
 
-   Run `reqstool status local -p <module.path>` and confirm the new requirement appears.
+   - **If MCP is available**: call `get_requirement_status(<new_req_id>)` on the `reqstool` MCP server and confirm the requirement is returned.
+   - **Otherwise**: run `reqstool status local -p <module.path>` and confirm the new requirement appears.
 
 7. **Report**
 

plugins/reqstool/skills/reqstool/add-svc/SKILL.md

@@ -94,10 +94,10 @@ For reqstool fundamentals (imports, filters, implementations), see `reqstool-ove
    Add the new SVC ID to the `filters.<urn>.svc_ids.includes` list in the relevant module's
    `<module.path>/software_verification_cases.yml`.
 
-7. **Verify with reqstool**
+7. **Verify the new SVC is tracked**
 
-   Run `reqstool status local -p <module.path>` and confirm the new SVC appears
-   under the linked requirement.
+   - **If MCP is available**: call `get_svc(<new_svc_id>)` on the `reqstool` MCP server and confirm the SVC is returned.
+   - **Otherwise**: run `reqstool status local -p <module.path>` and confirm the new SVC appears under the linked requirement.
 
 8. **Report**
 

plugins/reqstool/skills/reqstool/init/SKILL.md

@@ -78,20 +78,51 @@ Create or update `.reqstool-ai.yaml` configuration for this project.
    structure, comments, and formatting. Replace placeholder values with the
    user's inputs and include only the modules the user defined.
 
-5. **Verify (if possible)**
+5. **Configure MCP server (optional)**
+
+   Offer to configure the reqstool MCP server (requires reqstool ≥ 0.9.0):
+
+   > The reqstool MCP server gives AI tools structured access to your requirements.
+   > Would you like me to add the MCP server configuration?
+
+   If yes, use **AskUserQuestion** to ask the scope:
+   - **Project** — `.mcp.json` in the project root (shared with the team via version control)
+   - **Global** — `~/.config/claude/mcp.json` (just for you, not committed)
+
+   Read the chosen file (create if missing) and add or update the `reqstool` entry:
+
+   ```json
+   {
+     "mcpServers": {
+       "reqstool": {
+         "command": "reqstool",
+         "args": ["mcp", "local", "-p", "<system.path>"]
+       }
+     }
+   }
+   ```
+
+   If the chosen file already has a `reqstool` entry, show it and ask before overwriting.
+
+   If `reqstool` is not yet installed, skip this step and tell the user:
+   > Install reqstool first (`pipx install reqstool`), then re-run `/reqstool:init`
+   > or add the MCP config manually.
+
+6. **Verify (if possible)**
 
    If `reqstool` CLI is installed **and** the system path contains existing reqstool files,
    run `reqstool status local -p <system.path>` to verify the config.
 
    If reqstool is not installed or files don't exist yet, skip verification and
    tell the user they can run `/reqstool:status` later once their reqstool files are in place.
 
-6. **Report**
+7. **Report**
 
    Show the user:
    - The created/updated config (print the file contents)
    - Whether this was a new file or an update
-   - Remind them to add `.reqstool-ai.yaml` to version control
+   - Whether the MCP server was configured and in which scope
+   - Remind them to add `.reqstool-ai.yaml` to version control (and `.mcp.json` if project-scoped)
    - Next steps: "Run `/reqstool:add-req` to add your first requirement,
      or `/reqstool:status` to check traceability status."
 

plugins/reqstool/skills/reqstool/status/SKILL.md

@@ -7,7 +7,7 @@ metadata:
   version: "1.0"
 ---
 
-Run `reqstool status local` for a reqstool directory in this project.
+Show requirements traceability status. Uses the reqstool MCP server if configured, falls back to the CLI.
 
 ---
 
@@ -30,7 +30,22 @@ Read `.reqstool-ai.yaml` — see `reqstool-conventions.md` for field reference.
    - If the user is working inside a module directory (e.g., a Gradle subproject),
      ask whether they want to run status for that module or for the system level.
 
-3. **Run reqstool status**
+3. **Try MCP first**
+
+   Call the `reqstool` MCP server `get_status` tool.
+
+   - If it succeeds: present the returned status (requirements met/total, test summary).
+     Note: the MCP server always reflects the system-level path it was started with.
+     If the user asked for a specific module and the MCP result does not scope to that path,
+     note this limitation and proceed to the CLI fallback for the module path.
+
+4. **Fall back to CLI**
+
+   If the MCP server is not configured or the call fails, immediately tell the user:
+
+   > The reqstool MCP server is not configured — falling back to CLI.
+
+   Then run:
 
    ```bash
    reqstool status local -p <path>
@@ -39,14 +54,15 @@ Read `.reqstool-ai.yaml` — see `reqstool-conventions.md` for field reference.
    Show the output directly — reqstool traverses imports and implementation
    configuration automatically, so no further summarization is needed.
 
-4. **If reqstool is not installed**
+5. **If neither works**
 
-   If the command fails with "not found", tell the user:
-   ```
-   reqstool is not installed. Install with: pipx install reqstool
-   ```
+   If `reqstool` CLI is also not found, tell the user:
+
+   > Both the reqstool MCP server and CLI are unavailable.
+   > - To install the CLI: `pipx install reqstool`
+   > - To configure the MCP server: run `/reqstool:init`
 
 **Guardrails**
 - Always run from the project root directory
 - Do not modify any files — this is a read-only status command
-- Do not summarize or reformat reqstool output — show it as-is
+- Do not summarize or reformat reqstool CLI output — show it as-is