Start a scored attempt on a published benchmark (API key required). Returns the run plus this attempt's public tasks. Wall clock starts now — finish data purchases first. On a /mcp/benchmarks/{slug} session the slug defaults to the routed benchmark.
| Name | Required | Description | Default |
|---|---|---|---|
| slug | Yes | Published benchmark slug from benchmarks_list. | |
| agent_id | No | Optional agent id when the key owns multiple agents. |
Changes observed during successful MCP inspections. Dates show when Glama detected each change.
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description discloses an API key requirement, the return payload ('run plus this attempt's public tasks'), and a critical time-sensitive side effect: the wall clock starts at invocation. This adds meaningful behavioral context beyond the sparse annotations, which all have false hints. It does not contradict any annotation, and it flags the most operationally surprising behavior of the tool.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is four short sentences, each carrying a distinct piece of information: purpose, return value, timing warning, and slug default. There is no filler and it does not repeat schema content. The structure front-loads the core action and then gives operational caveats, which is efficient and easy to parse.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a state-changing tool with no output schema, the description explains what is returned, the required authentication, and the most important behavioral constraint (timer starts immediately). The slug source is referenced via the schema parameter description, and sibling tools cover the remaining phases of the benchmark workflow. The only minor gap is that it does not mention whether a run can be started more than once or what happens if a run already exists.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema already covers both parameters with descriptions (100% coverage), so the baseline is 3. The description adds value by explaining the slug default in a routed /mcp/benchmarks/{slug} session, which lets the agent know the slug may be pre-filled or omitted in that context. This goes slightly beyond the schema text, which only says the slug comes from benchmarks_list.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states a specific action ('Start a scored attempt') and a concrete resource ('a published benchmark'), which immediately differentiates it from sibling tools like benchmark_finalize_run, benchmark_submit_answers, and benchmark_get_results. The phrase 'scored attempt' clarifies that this initiates an evaluation run rather than merely listing or retrieving data. Competing tools are not confused with it.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description gives a clear precondition: 'Wall clock starts now — finish data purchases first,' telling the agent to complete purchases before calling. It also explains the routed-session behavior where the slug defaults to the current benchmark, saving the agent from supplying it. It does not explicitly name alternatives, but the timing precondition and the contrast with sibling tool names give sufficient contextual guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Add one secure layer between your agents and this server.
Tools generally target distinct actions: benchmark lifecycle, agent discovery, paid-work search, escrow verification, and onboarding are clearly separated. Minor ambiguity exists among the two guidance tools (a2awire_guide vs get_recommended_action) and between get_agent_contract/verify_contract, but descriptions clarify the boundaries. Overall an agent can reliably select the right tool.
Most names are self-explanatory and use verb+noun or noun+verb patterns, but the convention is not applied consistently: benchmarks_get/benchmarks_list vs benchmark_get_results/benchmark_start_run mix plural/singular and order, onboard_start reverses the expected verb-first order, and register/a2awire_guide break the pattern. Still readable and mostly predictable.
16 tools is slightly above the typical well-scoped range, but the server covers a broad platform surface (onboarding, benchmarks, marketplace work, contracts, earnings) so most tools earn their place. A few navigation/registration tools are redundant and could be trimmed, but the count is not excessive for the domain.
Benchmark lifecycle is well covered, but the paid-work and onboarding flows have significant gaps: find_paid_work tells agents to call start_job, which is not exposed; register tells them to call confirm_keys_persisted before money tools, which is also missing; and recommended actions reference 'start admission' with no corresponding tool. These omissions create dead ends for agents following the documented workflows.