An MCP server that recursively spawns Goose instances, allowing agents to use Goose to solve complex problems.
Mother Goose provides 8 tools with mostly complete schemas and descriptions, but has notable gaps in parameter documentation and error handling guidance. All tools have basic descriptions (10-50 chars), but many parameter descriptions are minimal or absent. Schemas are present and properly typed for required parameters. The server manages a complex domain (recursive Goose process management) but lacks LLM-optimized descriptions and missing error recovery guidance. Naming is verb-forward and mostly clear, but some descriptions are too terse to guide LLM selection. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are present despite clear safety distinctions (WRITE, READ_ONLY, DESTRUCTIVE risk levels in metadata).
Configure the circuit breaker settings to prevent token runaway
Get the current output from a specific gosling process with pagination support
Get a compact activity status report of goslings with their working/idle state
List all running and completed Goose processes
Release a specific gosling process when you're done with it
Run a Goose command with the specified prompt
Send a follow-up prompt to a running gosling process
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear safety profiles. run_goose, send_prompt_to_gosling, configure_circuit_breaker, and terminate_all_goslings are marked WRITE/DESTRUCTIVE in metadata but annotations are not exposed in tool definitions.
Output schemas are not documented for any tool. LLMs cannot plan downstream tool calls or understand what data will be returned. For example, list_goslings returns gosling objects but the structure (fields, types) is invisible to the agent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 63 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 28 | - | v1 |
Terminate all running gosling processes
terminate_all_goslings lacks a confirmation/dry-run pattern. This is a destructive operation that will terminate all running processes, but there is no confirmation step or warning in the description. Agents may invoke it unintentionally.
Error handling guidance is absent across all tools. Descriptions do not explain failure modes (e.g., 'If gosling already terminated, returns error X, try list_goslings() to check status'). LLMs cannot self-correct or recover gracefully.
Parameter bounds are missing for numeric parameters. max_active_goslings, max_total_goslings, max_runtime_minutes, max_output_size_bytes, max_prompts_per_gosling, auto_terminate_idle_minutes, limit, offset, idle_threshold_ms lack min/max constraints. LLMs may pass invalid or absurd values.
Most descriptions are 10-60 chars and lack context on when to use each tool vs. alternatives (e.g., distinction between list_goslings and get_gosling_status).