docs(plan): mark mcp transport plan completed

Author: rexleimo

docs/plans/2026-03-11-mcp-transport.md

@@ -1,13 +1,23 @@
 # MCP Transport (stdio) Implementation Plan
 
-> **For agentic workers:** REQUIRED: Use superpowers:subagent-driven-development (if available) or superpowers:executing-plans to implement this plan. Steps use checkbox (`- [ ]`) syntax for tracking.
+> **For agentic workers:** REQUIRED: Use superpowers:subagent-driven-development (if available) or superpowers:executing-plans to implement this plan. Steps use markdown checkboxes for tracking.
 
 **Goal:** Add MCP client support (stdio transport) so LoopForge can load an `mcp-servers.json`, expose MCP tools to the model, and route tool calls/resources/prompts through MCP.
 
 **Architecture:** Session-scoped MCP config is stored in runtime KV. Toolset owns MCP transports and exposes namespaced tool definitions (`mcp_<server>__<tool>`) plus wrapper tools for resources/prompts.
 
 **Tech Stack:** Rust, tokio, serde_json, newline-delimited JSON-RPC over stdio.
 
+**Execution Status (2026-04-07):** Completed and merged on `main` (initial landing: `f33803e`; follow-up hardening and coverage included up to `0609812`).
+
+**Re-Verification (2026-04-07):**
+- `cargo test -p loopforge-cli --locked cli::tests -q` ✅
+- `cargo test -p rexos-tools --locked mcp -q` ✅
+- `cargo test -p rexos-tools --locked -q` ✅
+- `cargo test -p rexos --locked agent_loop -q` ✅
+- `python3 -m mkdocs build --strict` ✅
+- `make check` ✅
+
 ---
 
 ## Chunk 1: CLI + session storage plumbing
@@ -20,20 +30,20 @@
 - Modify: `crates/rexos-runtime/src/session_skills/storage.rs`
 - Test: `crates/loopforge-cli/src/cli/tests.rs` (CLI parse)
 
-- [ ] **Step 1: Add `--mcp-config` to `loopforge agent run`**
+- [x] **Step 1: Add `--mcp-config` to `loopforge agent run`**
   - Add `mcp_config: Option<PathBuf>` argument.
 
-- [ ] **Step 2: CLI reads JSON file and stores content for the session**
+- [x] **Step 2: CLI reads JSON file and stores content for the session**
   - Read file bytes, parse JSON for validity, store normalized string in runtime session KV.
 
-- [ ] **Step 3: Add runtime APIs**
+- [x] **Step 3: Add runtime APIs**
   - `set_session_mcp_config(session_id, raw_json)`
   - `load_session_mcp_config(session_id) -> Option<String>`
 
-- [ ] **Step 4: Update CLI tests**
+- [x] **Step 4: Update CLI tests**
   - Ensure the new flag parses.
 
-- [ ] **Step 5: Run focused tests**
+- [x] **Step 5: Run focused tests**
   - Run: `cargo test -p loopforge-cli cli::tests -q`
   - Expected: PASS
 
@@ -51,24 +61,24 @@
 - Create: `crates/rexos-tools/src/mcp/types.rs`
 - Test: `crates/rexos-tools/src/mcp/tests.rs`
 
-- [ ] **Step 1: Define config types**
+- [x] **Step 1: Define config types**
   - `McpServersConfig { servers: BTreeMap<String, McpServerConfig> }`
   - `McpServerConfig { command, args, env, cwd }`
 
-- [ ] **Step 2: Implement JSON-RPC client helper**
+- [x] **Step 2: Implement JSON-RPC client helper**
   - `send_request(method, params) -> Result<Value>`
   - background read loop, oneshot per id, timeout support.
 
-- [ ] **Step 3: Implement stdio transport**
+- [x] **Step 3: Implement stdio transport**
   - spawn child, newline-delimited JSON, capture stderr tail on failure.
 
-- [ ] **Step 4: Implement MCP handshake**
+- [x] **Step 4: Implement MCP handshake**
   - `initialize` then `initialized` notification.
 
-- [ ] **Step 5: Unit tests for JSON-RPC routing**
+- [x] **Step 5: Unit tests for JSON-RPC routing**
   - use a minimal in-process “fake transport” (or spawn stub binary) to validate request/response matching.
 
-- [ ] **Step 6: Run focused tests**
+- [x] **Step 6: Run focused tests**
   - Run: `cargo test -p rexos-tools mcp -q`
   - Expected: PASS
 
@@ -92,20 +102,20 @@
 - Create: `crates/rexos-tools/src/dispatch/mcp/mod.rs`
 - Test: `crates/rexos-tools/src/tests/compat.rs` (definitions include MCP wrappers when enabled)
 
-- [ ] **Step 1: Add MCP wrapper tool defs**
+- [x] **Step 1: Add MCP wrapper tool defs**
   - `mcp_resources_list`, `mcp_resources_read`, `mcp_prompts_list`, `mcp_prompts_get`, `mcp_servers_list`
 
-- [ ] **Step 2: Extend Toolset with optional MCP hub**
+- [x] **Step 2: Extend Toolset with optional MCP hub**
   - On build: parse raw JSON from runtime, initialize servers, list tools.
 
-- [ ] **Step 3: Flatten remote MCP tools into `ToolDefinition`s**
+- [x] **Step 3: Flatten remote MCP tools into `ToolDefinition`s**
   - name: `mcp_<server>__<tool>`
   - parameters: remote `inputSchema` (default `{ "type": "object" }`)
 
-- [ ] **Step 4: Add dispatch routing**
+- [x] **Step 4: Add dispatch routing**
   - New domain `ToolCallDomain::Mcp` and route calls to MCP hub.
 
-- [ ] **Step 5: Run focused tests**
+- [x] **Step 5: Run focused tests**
   - Run: `cargo test -p rexos-tools -q`
   - Expected: PASS
 
@@ -120,16 +130,16 @@
 - Modify: `crates/rexos-runtime/src/session_skills/storage.rs`
 - Test: `crates/rexos/tests/agent_loop.rs` (add a small case with MCP stub)
 
-- [ ] **Step 1: Load session MCP config in `run_session`**
+- [x] **Step 1: Load session MCP config in `run_session`**
   - If present, build Toolset with MCP enabled.
 
-- [ ] **Step 2: Ensure tool defs include MCP tools**
+- [x] **Step 2: Ensure tool defs include MCP tools**
   - Toolset initializes MCP before `tool_defs` is computed.
 
-- [ ] **Step 3: Add an integration-ish test**
+- [x] **Step 3: Add an integration-ish test**
   - Spawn MCP stdio stub, configure via `--mcp-config`-equivalent session KV, verify a tool call succeeds.
 
-- [ ] **Step 4: Run runtime tests**
+- [x] **Step 4: Run runtime tests**
   - Run: `cargo test -p rexos agent_loop -q`
   - Expected: PASS
 
@@ -143,15 +153,15 @@
 - Modify: `docs-site/blog/mcp-integration-guide.md`
 - Modify: `docs-site/reference/config.md` (optional: mention session flag, not TOML)
 
-- [ ] **Step 1: Update MCP blog to match real CLI**
+- [x] **Step 1: Update MCP blog to match real CLI**
   - Keep `--mcp-config` example, remove commands that do not exist (`config add-mcp-server`).
 
-- [ ] **Step 2: Verify docs build**
+- [x] **Step 2: Verify docs build**
   - Run: `python3 -m mkdocs build --strict`
   - Expected: PASS
 
 ### Task 6: Full verification
 
-- [ ] **Step 1: Run full checks**
+- [x] **Step 1: Run full checks**
   - Run: `make check`
   - Expected: PASS