One API. The right route for each task.
MUSCLE Flex matches each supported coding-agent task to the best qualified model path for that work. Your integration stays fixed while the frontier keeps moving.
Already invited? Sign in to the portal
One request, start to finish
Checks first. Then the route.
Every request passes capability and funding checks before any route is chosen. The response comes back in the shape your SDK expects, with a request ID for support.
- POST
- /v1/responses
- model
- muscle/auto
- tools
- 3
- stream
- true
- capability check
- pass
- funding admission
- pass
- route
- selected for the task
- tool call
- inspect_repository
- continuation
- preserved
- usage
- settled
- request
- req_••••7K2Q
Different tasks need different kinds of intelligence.
A fast explanation, a tool-heavy repository search, and a multi-file repair reward different model strengths. A production agent may also stream, call tools, continue for many turns, and cross several cost boundaries. Flex evaluates the route around the task and the workflow, not a one-line benchmark prompt.
The best route starts with qualification.
Flex first excludes routes that cannot meet the request’s hard requirements: the supported tool shape, stream semantics, context needs, account policy, and available funding. Only then does the routing policy compare the eligible paths and choose the best qualified fit for the task.
A low score cannot make an incompatible route safe. Hard requirements stay hard.
Your application keeps control of the tools.
Flex returns function calls in the supported API shape. Your application decides whether to run them, runs them in its own environment, and sends the results back. Flex never gains shell access or takes over the repository because it routed the model call.
Prepaid usage with a hard stop.
Beta accounts fund a service balance before use. Flex holds enough balance before admitting new paid work and settles measured usage afterward. When the available balance is too low, new work is declined. The portal shows the remaining balance and recent activity.
New models earn their place.
Flex does not route your work to a model because it is new or popular. Candidate paths are tested with controlled coding tasks, real SDK calls, tool continuations, concurrent requests, cancellation and transport cases, funding denial, and accounting replay. They stay out of production until the relevant gates pass.
Similar wire shapes. Explicit boundaries.
Flex supports selected OpenAI-shaped and Anthropic-shaped workflows. It does not claim every endpoint, field, hosted tool, media type, or background mode. The compatibility matrix is the contract: each behavior is marked PASS, FAIL, EXCLUDED, or Pending.