How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

Workbench - Skill - OpenSmartRoute

Imported from mcunille/Workbench (AGENTS.md). Install upstream with npx skills add mcunille/Workbench. Copyright stays with the author.

Repository Guidance

Workflow and authority

Design, implementation, review: apply the five-step algorithm in order. Changing safeguards requires the applicable approval.
Before implementation, complete the applicable superpowers design and planning handoffs. Bounded changes require short design approval; approval covers the presented artifact, not later unseen artifacts.
After verified implementation, commit scoped changes, push the branch, and open or update a ready-for-review PR against its base branch. This preselects superpowers:finishing-a-development-branch's PR option; skip the integration-menu question. Create a draft only when requested.
Preserve skills' separate approval gates for review publication and other collaboration writes. Delivery approval does not authorize those writes. Merge and production operations require explicit user authorization.
Keep incidental cleanup small, directly related, and low-risk. Surface broader refactors and design tradeoffs separately with evidence and affected locations; obtain the required approval before filing issues. Preserve unrelated working-tree changes, and inspect logs and diffs before concluding work in another checkout is safe to discard.
Preserve other checkouts' resources. Keep credential values out of chat, logs, command arguments, test output, diffs, generated artifacts, and Git history.

Task routes

Execution, debugging, completion claims: follow execution and verification.
Module design, internal review, PR review, feedback, UI work, or test authoring: follow skill routing. For codebase-design, review-pr, or handle-pr-feedback, also follow installed skill ownership.
Documentation audits or cleanup: use doc-audit and read audit scope, bookkeeping, and skill ownership.
Behavior changes, refactors, test edits, or verification: follow test method, including current-source evidence. Documentation-only and formatting-only changes use its affected-check guidance.
Application changes or API generation: follow CONTRIBUTING.md. Architecture or product contracts: locate the owner through docs/README.md.
Preview lifecycle: follow setup. Credentials: follow credential use; existing .env.dev installations use manual development.
Schema changes: follow migration authoring and validation, including pre-PR consolidation and preview-data preservation.
Visual PR evidence: follow verification evidence.
Before executing or modifying checkout scripts, tool configuration, Impeccable, or hooks: review the developer tooling trust boundary.

Install Workbench and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

Workbench

About

Repository Guidance

Workflow and authority

Task routes

Install Workbench and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews