Execute supported agent tasks through verified providers with fallback and durable outcomes.
Listed on
- Official MCP Registry
- Smithery
- Glamavia MCP Toplist
- mcp.sovia MCP Toplist
First seen 2 Oct 2026. One server, whatever directories list it: each directory listing keeps its own page and history.
4
Directories
2 via MCP Toplist
4
Tools
From an anonymous probe
-
ToolBench grade
Not graded by Arcade
1
GitHub stars
From MCP Toplist
Tools
| Tool | Description | Behaviour |
|---|---|---|
| atl_complete_task | Complete a supported task end-to-end. This is ATL's default entry point for search, extract, summarize, translate, and classify when no provider/API/tool is explicitly fixed: ATL selects an eligible route, executes it, may use bounded fallback, verifies the result, records the Outcome, and returns the result. Use routing-only atl_decide only when the caller explicitly wants to inspect/control the route before execution. Public first use requires no ATL API key. Calls are subject to bounded timeouts and server-side rate/concurrency protections; failures return ok=false with execution/result_validation evidence. | Changes data |
| atl_decide | Inspect routing without executing provider work. Use this only when the caller wants an ATL Decision, route evidence, or a decision-bound handoff before execution; for normal task-in/result-out work use atl_complete_task. ATL returns a structured Decision with selected route, reason_codes, Outcome correlation data, and next_step; if execution is later desired, pass the exact decision_reference and outcome_correlation_token from that handoff to atl_execute. Public first use requires no ATL API key. Calls are subject to bounded timeout and server-side rate/concurrency protections; invalid constraints or unavailable routes return structured decision/error evidence. | Read-only |
| atl_execute | Execute an existing ATL Decision; this tool does not choose a new route. It requires the exact decision_reference and outcome_correlation_token returned by atl_decide/next_step. ATL validates the Decision, executes the selected provider/tool, may use bounded fallback, and records Outcome automatically. Use only after a routing-only flow; for normal task-in/result-out work use atl_complete_task. Provider authentication requirements come from the Decision. On success structuredContent reports ok=true with execution_status, output_text, provider/attempt, latency/cost, and Outcome evidence; failures set ok=false with structured error evidence. Calls are subject to bounded timeouts/retries and server-side rate/concurrency protections. | Changes data |
| atl_outcome | Record external execution evidence for an existing ATL Decision; this tool never executes a provider. Use it only when the caller performed execution outside ATL. Send outcome_status with the exact decision_reference and outcome_correlation_token returned by ATL; add failure fields only when observed and cost/latency only when known. Duplicate reports are deduplicated, and invalid or mismatched correlation is rejected; accepted evidence can improve future routing/reputation. Do not call this after atl_complete_task or atl_execute because ATL-owned execution records Outcome automatically. The response reports ok, accepted, duplicate, status_code, and feedback/attribution evidence. Calls are subject to server-side rate/concurrency protections. | Changes data |
| Directory | Listing | Tier | First seen |
|---|---|---|---|
| Official MCP Registry | Agent Traffic Lab | - | 2 Oct 2026 |
| Smithery | Agent Traffic Lab | - | 2 Oct 2026 |
| Glama | Listed there according to MCP Toplist’s dataset; not collected by InvokeRank. | ||
| mcp.so | Listed there according to MCP Toplist’s dataset; not collected by InvokeRank. | ||