Rhumb Resolve
One key, many superpowers.
Give agents one governed execution surface for supported work. Start with Rhumb-managed capabilities, then bring your systems only when the workflow actually needs them.
Rhumb Resolve is Rhumb’s governed execution layer for AI agents: one integration surface for supported capabilities, with managed credentials, explainable routing, optional provider pinning, fallback where a supported alternate is configured, budget controls, and pay-per-call pricing.
Resolve routes each call to the best-fit provider for the call by default, while keeping explicit provider choice available when the agent wants direct control. One key, many superpowers.
Tell Resolve what you want accomplished, or use the supported provider/tool path you want in the way you want when direct control is better. The product story is one surface, not a tool graveyard.
Today Resolve is strongest in research, extraction, generation, and narrow enrichment across 28 callable providers. Discovery is broader than current execution coverage, and Rhumb keeps that boundary explicit.
First executable proof
Try the no-auth readiness check before execution.
curl -sS -H "X-Rhumb-Client: web" "https://api.rhumb.dev/v1/capabilities/email.send/resolve"email.send before any side effect. Use it to inspect the route, credential modes, circuit state, and execute readiness before choosing a paid execution rail.
For the narrower web-search pilot, follow the three-call search.query path: resolve, estimate, then execute only with a governed key or payment rail.
Best-fit routing
Index publishes the AN Score. Resolve first matches the supported capability path, then routes the call using AN Score as a major input alongside provider availability / circuit state, estimated cost, credential mode, latency proxy, and explicit constraints.
Intent-first or explicit control
Say what job you want done and let Resolve route it, or explicitly pin the supported provider/tool path when direct control is the point.
Honest execution boundary
Resolve is launchable today for research, extraction, generation, and narrow enrichment across 28 callable providers. It does not pretend every scored service is executable.
Before you wire this deeper
Want to inspect the trust story first? Read Trust, inspect the public Methodology, or use the provider dispute path if a score, routing claim, or coverage claim looks wrong.
Resolve authority map
The supporting pages are now first-class product truth.
Start with the overview, inspect the route proof, compare adjacent platforms, then choose the right credential and pricing rail.
What is Resolve?
The plain-language system model: Index ranks, Resolve routes, and execution stays governed.
How Resolve routes calls
Route factors, AN Score role, explicit provider pinning, receipts, and neutrality/cost disclosure.
Resolve comparisons
When to use Rhumb versus connector catalogs, tool authorization layers, and OAuth infrastructure.
Keys and pricing
Credential paths, governed API keys, BYOK, Agent Vault, x402, and per-call economics.
What Resolve is
Governed execution for agents that need real work done.
Agents can ask Resolve for a capability or explicitly pin the supported provider path they want. By default, Resolve chooses from providers mapped to that capability, using AN Score as a major input alongside provider availability / circuit state, estimated cost, credential mode, latency proxy, and explicit policy constraints.
Start with Rhumb-managed capabilities where Rhumb can honestly own the execution path. Bring BYOK or Agent Vault only when the workflow touches your own systems, permissions, or compliance boundary.
The product truth is simple: Index ranks. Resolve routes.
And agent control stays first-class: say what you want done, or pin the supported provider path yourself when that is the better way to work.
How it works
01. Index ranks the field
Rhumb Index scores services for AI-agent compatibility and gives Resolve a public quality prior.
02. Resolve chooses the route — unless you do
Resolve picks the best-fit supported provider for the call by default instead of blindly following the global leaderboard, while still allowing explicit provider pinning when the agent wants direct control.
03. You see the operating truth
Estimate the route, cost, and readiness before execution so the agent can explain the path it took.
What decides the route?
Task-aligned routing, not blind leaderboard routing.
Resolve does not blindly pick the global top score. It picks the best-fit supported provider for the call by default, while still allowing explicit provider choice when the agent wants direct control.
Route by supported capability, runtime factors, and explicit constraints — not leaderboard purity. AN Score stays central, but it works as a quality prior inside an explainable routing system.
Read how Resolve routes calls →AN Score
Rhumb’s public quality prior. Important, but not the only routing input.
Capability fit
Which supported provider is actually strongest for the requested job.
Availability
Whether the provider route is available enough to trust for this execution.
Cost
Estimated per-call cost can constrain routing without rewriting the independent AN Score.
Credential mode
Rhumb-managed, BYOK, and Agent Vault create different honest execution paths.
Constraints
Budget ceilings, allow/deny lists, and explicit pins shape the route when they materially matter.
Current launchable surface
Broad enough to matter. Narrow enough to stay honest.
Current launchable scope: research, extraction, generation, and narrow enrichment — not general business-agent automation.
Not every service or capability in the index is executable through Rhumb today. Discovery breadth is wider than current callable coverage.
Examples
When supported routing beats raw rank
Fresh social / live context
When the task depends on current X-native information or real user-input-sensitive context, Resolve can prefer the provider that is strongest for that job instead of blindly following the generic overall ranking.
Web extraction
When structure and extraction quality matter more than generic ranking, Resolve can route to the provider that best fits extraction work.
Generation / transform
When the task is generation or transformation, Resolve routes based on the real capability path, not just the broad service score.
Credential paths
Three credential paths, one trust story
Resolve should not force every capability through the same trust model. Rhumb-managed, BYOK, and Agent Vault each exist for a reason.
Default path
Rhumb-managed
Best for zero-config utility capabilities where Rhumb can honestly own the provider account, route the call, and expose the capability as the product.
Your systems
BYOK
Use when the workflow touches your account data, workspace, or production systems and you want direct control over the upstream credential.
Agent-native custody
Agent Vault
Use when the credential should stay encrypted in Rhumb custody, scoped to the agent, and be injected only at execution time.
FAQ
Resolve, plainly
Does Resolve just follow the top score?
No. AN Score is a major routing input, not the only one. Resolve chooses the best-fit supported provider for the call by first matching the supported capability path, then using AN Score, provider availability / circuit state, estimated cost, credential mode, latency proxy, and explicit constraints.
Can I pick the provider or tool path myself?
Yes. Rhumb supports both intent-first routing and explicit provider choice. Ask for the job and let Resolve route it, or pin the provider yourself on the raw-provider surface when direct control is the point.
What is executable today?
Current launchable scope: research, extraction, generation, and narrow enrichment — not general business-agent automation. Resolve is strongest today in research, extraction, generation, and narrow enrichment across 28 callable providers.
When should I bring my own systems?
Start with Rhumb-managed capabilities where Rhumb can honestly own the execution path. Bring BYOK or Agent Vault only when the workflow touches your own systems, permissions, or compliance boundary.