Loading ToolBench
An MCP server that exposes Apify Actors as tools, enabling AI agents to run web scraping, data processing, and automation workflows.
The server has 7 well-named tools with clear descriptions and explicit input schemas. All tools start with action verbs (call-, get-, list-, search-, report-) following convention. Most parameter descriptions are present and adequately detailed. However, output schemas are not documented in the source code provided, and error handling guidance is minimal. Tool descriptions average ~90 characters, which is somewhat short for LLM optimization (baseline is 194 chars). The schema documentation is present but lacks detail on response structure. Two tools (call-actor, report-problem) are write operations but descriptions could be more explicit about side effects and idempotency.
Execute an Apify Actor and retrieve its results
Retrieve detailed information about an Actor including input schema and description
Retrieve details about a specific Actor run including its status and results
List all runs for a specific Actor with optional filtering and pagination
List available Actors in the Apify platform
Report a problem with an Actor run or the MCP server to Apify support
Search for Actors by name or keyword using the Apify platform search
Output schemas not documented in source code. LLMs cannot infer return structure, list-actors, list-actor-runs, search-actors return arrays but field names and types are not visible in schema definitions provided.
Tool descriptions are brief (range ~55-70 chars) and lack WHEN-to-use guidance. LLMs need context on when to select call-actor vs get-actor-run, or when list-actors suffices vs search-actors. Baseline production tools average 194 chars.
call-actor (write operation) description does not explicitly state it modifies state or warn about side effects. Agents need to know this is irreversible. Similarly, report-problem (write) lacks clarity on side effects.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 64 | 2026-07-28+ | v2 |
| 2026-03-09 | B | 71 | 1.25.2+ | v1 |
Pagination parameters (limit, offset) are present on list and search tools, but no documentation of total count, has_more flag, or next_cursor is visible. Agents cannot determine if there are more results without explicit response fields.
No error recovery guidance in any tool description. If call-actor times out or actor fails, what should the LLM do? Retry? Call get-actor-run? This is not documented.