Name the job, not the vendor. Routes a provider, fails over, catalog off the tool list.
Scored across 7 tools
Each tool maps to a distinct task: capability-level execution, tool-id execution, multi-provider comparison, discovery, reliability stats, vendor API proxying, and billing. The potential overlaps (act vs call vs compare) are clearly separated by failover, fixed-tool vs multi-provider behavior.
All names share a clean thruact_ prefix and lowercase single-word convention, which is easy to predict. The suffix is not strictly verb_noun because iq, proxy, and walet are noun-ish while act, call, compare, and discover are verbs, so there is a minor stylistic inconsistency.
Seven tools is well within the ideal range and each one has a clear purpose with no redundant duplication. The count matches a gateway-style product that needs discovery, execution modes, comparison, reliability, proxying, and wallet visibility.
The set covers the main user-facing workflow: discover capabilities, execute them, compare answers, inspect reliability, proxy to external vendors, and track spending. The only noticeable gap is that there is no explicit directory/list tool for connected vendor ids or all available tool ids, though discover and external workspace setup can partially work around that.
Run a capability. Routes across providers and fails over automatically on provider faults. The reply's route says which providers were considered, what each scored, and why any were not used — including ones dropped for a missing credential.
Inputs: capability (required), hedge, input, prefer, +1 more.
Run one specific tool by id. No failover.
Inputs: tool (required), params.
Ask every provider behind a capability the same question at once and report where their answers differ. Use it when the answer matters enough to want it corroborated. Costs more than act: every provider called is paid.
Inputs: capability (required), input, providers.
Find capabilities for a job, with prices and measured reliability. Search first, execute second.
Inputs: query (required), limit.
Measured reliability per tool over the last 7 days.
Inputs: capability.
Send an authenticated request to a third-party vendor the workspace has connected. The stored API key is injected server-side — you never see it. Give the vendor id from the directory (e.g. openai, sendgrid), an HTTP method, and a path on that vendor's own API. Use it to act through a connected vendor that is not a Thruact capability. Metered per call.
Inputs: path (required), vendor (required), body, method.
Prepaid balance and recent charges.
Inputs: no named parameters.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
OpenRouter for tools and data. Compare catalog providers and call them from one hosted MCP endpoint.
Find the right tool for any need across every family and imported API, then call it by name
Execute supported agent tasks through verified providers with fallback and durable outcomes.
Rank MCP servers/APIs for a job. Lexical ranking, 36 curated tools, no Stripe.
Add one secure layer between your agents and this server.
