Caller-specific static MCP capability fit evidence. No live probing; caller owns the decision.
Scored across 1 tool
There is only one tool, so no selection ambiguity is possible. Its purpose (checking whether a capability revision satisfies declared workflow requirements) is stated precisely and has no competing sibling.
check_mcp_capability follows a clean verb_noun snake_case pattern. With a single tool there is no opportunity for convention drift.
One tool is thin for a server framed as a preflight capability, even if the domain is narrow. The description itself references adjacent operations (POST /v0/snapshot-lint, decision attestation) that are not exposed as tools.
The check itself is self-contained, but the surface has real gaps: no tool to lint/validate arguments without a credential, no tool to create or capture a capability snapshot, and no tool to attest a decision afterward, all of which the description says are required parts of the workflow.
Before relying on an MCP capability, answer one question about it: does this exact revision satisfy the requirements your workflow declares? Call it for new work, after a dependency changes, after your own requirements change, or when an integration has already failed and you need to know whether the capability is why (the public triggers are NEW_WORK, DEPENDENCY_CHANGE, REQUIREMENTS_CHANGE, INTEGRATION_FAILURE, OTHER_REAL_WORK). You supply both sides: a capability snapshot you already hold and your workflow requirements. Nothing is fetched and no endpoint is probed, so the answer is reproducible from what you sent and is only ever as good as that snapshot. You get back a verdict with per-check results, coverage, and the limits of the check itself. That is evidence, not an instruction: the caller owns CONTINUE, UPDATE_INTEGRATION, and HOLD, and no verdict maps to any of them. Optional previous (an earlier snapshot and its finished report) adds a revision_receipt comparing that conclusion with this revision. Budget before calling: a Preview grant allows three checks, while POST /v0/snapshot-lint validates argument and requirements shape with no credential and no quota, so start there when unsure. This tool never writes usage-events (usage_event_created is always false) and never discloses the internal Preview bearer; attest your own decision afterwards on a finished owned check.
| Name | Required | Description | Default |
|---|---|---|---|
| previous | No | Optional. Makes this a revision-aware check (POST /v0/revision-checks semantics). Send the snapshot and finished report from an earlier check of the same capability_key with identical requirements. Exactlane re-evaluates them instead of trusting the report and never looks up earlier calls. Omit for the ordinary check. | |
| snapshot | Yes | Caller-provided MCP capability snapshot. No live endpoint is accepted. | |
| requirements | Yes | Caller workflow requirements using a supported mcp-consumer-fit schema | |
| capability_key | Yes | Opaque caller capability key; same contract as POST /v0/capabilities | |
| revision_label | No | Optional revision label; defaults to mcp-adapter | |
| profile_version | No | Optional mcp-consumer-fit version; inferred from requirements when omitted |
| Name | Required | Description |
|---|---|---|
| report | Yes | Finished Exactlane check report, substantially unchanged |
| decision_note | No | |
| revision_receipt | No | Present only when previous was supplied: exactlane.revision-receipt/0.1. The global previous-to-current comparison is authoritative; isolated change evaluations report DIRECT_EFFECT_OBSERVED, NO_ISOLATED_EFFECT_OBSERVED or NOT_EVALUATED and cannot rule out interaction effects. |
| usage_event_created | Yes | Always false. check_mcp_capability does not POST /v0/usage-events and does not disclose the internal Preview bearer. To attest a client-owned CONTINUE, UPDATE_INTEGRATION, or HOLD decision, use REST POST /v0/usage-events after a finished owned check; report_consumed is required. |
| caller_owns_operational_decision | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Beyond the annotations it discloses reproducibility guarantees ('nothing is fetched and no endpoint is probed'), quota limits (three checks per Preview grant), the absence of usage-event writes, and that no verdict maps to an action. There is mild tension with readOnlyHint=false since the text emphasizes side-effect-freedom, but the claim is narrow ('never writes usage-events') rather than a blanket denial, so it is not a contradiction.
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 purpose is front-loaded, but the body is a dense wall of sentences packed with parenthetical asides and raw endpoint references, making key operational facts (quota, alternative endpoint, no-write guarantee) hard to scan. Most sentences carry content, but the lack of paragraphing hurts structure.
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 complex tool with nested objects, six parameters, and an output schema, the description covers triggers, inputs, verdict shape, limits, and quotas adequately; return values are correctly left to the output schema. It never defines what a 'supported mcp-consumer-fit schema' is, leaving one input contract underspecified.
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?
Schema description coverage is 100%, so the baseline is 3, but the description adds real meaning: it explains that the caller supplies both sides (snapshot + requirements) and that optional 'previous' converts the call into a revision-aware check that yields a revision_receipt comparing conclusions. revision_label and profile_version are left to the schema.
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 opening sentence states a specific operation on a specific resource: evaluate whether an exact revision satisfies declared workflow requirements, and it explicitly frames the answer as a verdict with per-check results. An agent knows this is a pure evaluation tool, not a mutation or retrieval operation.
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?
It enumerates exact call triggers (NEW_WORK, DEPENDENCY_CHANGE, REQUIREMENTS_CHANGE, INTEGRATION_FAILURE, OTHER_REAL_WORK) and names an alternative path for the uncertain case (POST /v0/snapshot-lint, no credential, no quota). Both when-to-use and a fallback are explicit.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
check_mcp_capabilityEvidence-linked MCP selection and non-executing Safe-Call review; surfaces unknowns.
Market-validated MCP capabilities with x402 paid execution.
The evidence layer for MCP: live operational grades plus Trust Receipts for every registry server.
Vouch — independently measured reliability scores for MCP tools, not self-reported claims.
Add one secure layer between your agents and this server.