Interactive MCP client that turns a server's surface into terminal commands
mcp-repl is a REPL client, not a production tool server. The 9 tools in examples/mcp_repl_fixture.rs are test fixtures for demonstrating the client's capabilities, not real-world tools. Tool descriptions are minimal (10-50 chars), parameter descriptions are absent, and schemas lack detail. Most tools have no documented output structure. This is appropriate for a fixture but would fail production standards. The server itself (tower-mcp 0.22.1, protocol-2026-07-28) is well-maintained, but the fixture tools are intentionally sparse.
Add two integers
Emit a deterministic server log notification
A server tool named builtin
Return a deterministic MCP tool error
Fail from inside a task
Report deterministic process environment for import tests
Add two integers in a final-protocol task
Generic tool names ('tool', 'builtin') lack action verbs and are non-descriptive. LLMs cannot infer intent from these names alone.
Most tools (announce, fail, fail_slowly, process_info) have no visible input schema. Cannot verify parameter types or constraints.
Descriptions are uniformly short (10-35 chars) and lack context on WHEN to use each tool or WHAT it returns. Baseline for A+ tools is 50-200 chars with clear purpose statements.
No parameter descriptions visible for any tool. LLMs cannot determine what 'a' and 'b' mean in 'add' without explicit documentation.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | <=2025-11-25 | v2 |
A server tool named tool
Wait in the fixture server
No output schemas documented for any tool. Agents cannot plan downstream calls or extract required fields without knowing return structure.