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.

repro - Prompt - OpenSmartRoute

Name: repro
Author: BPForbes

Prompt file imported from BPForbes/Homework-Central (.cursor/commands/repro.md). Copyright stays with the author.

/repro

Also: reproduce, write a repro.

Reproduce the failure before declaring a cause.

Recreate it with exact commands, inputs, and exit codes.
Write the repro to .cursor/thoughts/non-finalized/repro-<topic>.md. Redact secrets first (thoughts-layout.md).
Files the repro creates in the project tree are process output and never land on the committed timeline. Prefer a throwaway clone (git clone --no-hardlinks . /tmp/probe); when the repro must run in this worktree, use a reserved lower-case name (a _scratch/ directory or a .scratch infix) and clean up before reporting. Undo a repro that edited a tracked file with git checkout -- <exact path>, never git checkout -- . or git stash — the worktree is shared.
Do not claim a root cause until the repro ran (or the environment cannot run it — then say so).
A successful repro or fix still does not authorize a push. Only QA may give the OK to push. Coders who change code must still run applicable CodeQL.
Spawn helpers with /create-subagent asynchronously when parallel repros help.
Use any installed / skill that fits (/browser-automation, /buildkite-*).

Catalog: .cursor/skills/devops-multi-agent-team/references/agent-commands.md.

# /repro Also: `reproduce`, `write a repro`. Reproduce the failure before declaring a cause. 1. Recreate it with exact commands, inputs, and exit codes. 2. Write the repro to `.cursor/thoughts/non-finalized/repro-<topic>.md`. Redact secrets first ([thoughts-layout.md](../skills/devops-multi-agent-team/references/thoughts-layout.md)). 3. Files the repro *creates* in the project tree are process output and never land on the committed timeline. Prefer a throwaway clone (`git clone --no-hardlinks . /tmp/probe`); when the repro must run in this worktree, use a reserved lower-case name (a `_scratch/` directory or a `.scratch` infix) and clean up before reporting. Undo a repro that edited a tracked file with `git checkout -- <exact path>`, never `git checkout -- .` or `git stash` — the worktree is shared. 4. Do not claim a root cause until the repro ran (or the environment cannot run it — then say so). 5. A successful repro or fix still does **not** authorize a push. **Only QA may give the OK to push.** Coders who change code must still run applicable CodeQL. 6. Spawn helpers with `/create-subagent` asynchronously when parallel repros help. 7. Use any installed `/` skill that fits (`/browser-automation`, `/buildkite-*`). Catalog: `.cursor/skills/devops-multi-agent-team/references/agent-commands.md`.

Install repro 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.

{ "prompt": "# /repro\n\nAlso: `reproduce`, `write a repro`.\n\nReproduce the failure before declaring a cause.\n\n1. Recreate it with exact commands, inputs, and exit codes.\n2. Write the repro to `.cursor/thoughts/non-finalized/repro-<topic>.md`.\n Redact secrets first ([thoughts-layout.md](../skills/devops-multi-agent-team/references/thoughts-layout.md)).\n3. Files the repro *creates* in the project tree are process output and\n never land on the committed timeline. Prefer a throwaway clone\n (`git clone --no-hardlinks . /tmp/probe`); when the repro must run in\n this worktree, use a reserved lower-case name (a `_scratch/`\n directory or a `.scratch` infix) and clean up before reporting. Undo\n a repro that edited a tracked file with `git checkout -- <exact path>`,\n never `git checkout -- .` or `git stash` — the worktree is shared.\n4. Do not claim a root cause until the repro ran (or the environment cannot run it — then say so).\n5. A successful repro or fix still does **not** authorize a push.\n **Only QA may give the OK to push.** Coders who change code must\n still run applicable CodeQL.\n6. Spawn helpers with `/create-subagent` asynchronously when parallel repros help.\n7. Use any installed `/` skill that fits (`/browser-automation`, `/buildkite-*`).\n\nCatalog: `.cursor/skills/devops-multi-agent-team/references/agent-commands.md`.", "variables": [], "notes": "Reproduce a failure with exact commands before declaring a cause. Notes stay in .cursor/thoughts/non-finalized/.", "metadata": { "source": { "provider": "github-cursor-commands", "repository": "https://github.com/BPForbes/Homework-Central", "path": ".cursor/commands/repro.md", "ref": "5478979d43bc4674034d76037049c2f0734d057c", "url": "https://github.com/BPForbes/Homework-Central/blob/5478979d43bc4674034d76037049c2f0734d057c/.cursor/commands/repro.md", "key": "BPForbes/Homework-Central/.cursor/commands/repro.md" } } }

repro

About

/repro

Use it

Install repro 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

Manifest

Reviews