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.

drive-pr - Prompt - OpenSmartRoute

Name: drive-pr
Author: kayleschmoyer

Prompt file imported from kayleschmoyer/Legacy (.cursor/commands/drive-pr.md). Copyright stays with the author.

Arguments: wherever <args> appears below, substitute the text the user typed after the slash command.

You are a PR driver. Pick up an existing PR and shepherd it all the way to merge.

Session mode: spawn Claude Code with the launcher's worktree checkbox checked. The session starts in an ephemeral .claude/worktrees/<name>/ worktree. If the PR's branch already has a worktree, activate it; otherwise check out the PR's branch in the launcher worktree (git checkout <branch>) or create a sibling worktree via worktree_attach from the main repo.

Load the playbook via .agents/skills/pr-lifecycle/SKILL.md, then .agents/skills/pr-lifecycle/references/drive.md for the full single-PR shepherding workflow.

Drive PR to merge:

Specific invocation: /drive-pr <N> or /drive-pr https://github.com/.../pull/<N> — pass the PR number or URL directly.

Parse the PR number from the arguments (number, URL, or #N)
Verify GitHub and Jira auth (github_auth_status, jira_auth_status)
Inspect the PR: github_view_pr, github_pr_batch_status, github_get_review_threads
Extract the Jira ticket key from the PR title or branch name and transition to "In Progress" if needed
Enter a worktree for the PR's branch (activate existing, checkout in launcher worktree, or create sibling via worktree_attach)
Fix blockers in priority order: conflicts (rebase) → CI failures → review comments → stale branch
Handle events as they arrive (CI fail, new review, conflicts, approved) — Claude Code surfaces these natively — keep handling events until the PR merges
Post-merge: transition Jira to "Done", clean up labels, clean up worktree
MCP gap audit (mandatory): run mcp-gap-audit Phase 1 to identify and ticket any MCP tooling gaps from this session

Guardrails: Always validate before pushing — make pc locally; in cloud/timeout-prone sessions run the seven pc-* chunks per cloud-agent-mode (never monolithic make pc). Cap review fix rounds at 2. Delegate to sub-skills (rebase, debugging, pr-lifecycle) — don't re-implement. Never stop after a single fix — the event loop is the core. Cloud agents report blockers instead of attempting code changes. Respect agent-in-progress labels set by other agents.

Install drive-pr 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.

> **Arguments**: wherever `<args>` appears below, substitute the text the user typed after the slash command. You are a **PR driver**. Pick up an existing PR and shepherd it all the way to merge. > **Session mode**: spawn Claude Code with the launcher's `worktree` checkbox checked. The session starts in an ephemeral `.claude/worktrees/<name>/` worktree. If the PR's branch already has a worktree, activate it; otherwise check out the PR's branch in the launcher worktree (`git checkout <branch>`) or create a sibling worktree via `worktree_attach` from the main repo. Load the playbook via `.agents/skills/pr-lifecycle/SKILL.md`, then `.agents/skills/pr-lifecycle/references/drive.md` for the full single-PR shepherding workflow. Drive PR to merge: <args> > Specific invocation: `/drive-pr <N>` or `/drive-pr https://github.com/.../pull/<N>` — pass the PR number or URL directly. 1. Parse the PR number from the arguments (number, URL, or `#N`) 2. Verify GitHub and Jira auth (`github_auth_status`, `jira_auth_status`) 3. Inspect the PR: `github_view_pr`, `github_pr_batch_status`, `github_get_review_threads` 4. Extract the Jira ticket key from the PR title or branch name and transition to "In Progress" if needed 5. Enter a worktree for the PR's branch (activate existing, checkout in launcher worktree, or create sibling via `worktree_attach`) 6. Fix blockers in priority order: conflicts (rebase) → CI failures → review comments → stale branch 7. Handle events as they arrive (CI fail, new review, conflicts, approved) — Claude Code surfaces these natively — keep handling events until the PR merges 8. Post-merge: transition Jira to "Done", clean up labels, clean up worktree 9. **MCP gap audit** (mandatory): run `mcp-gap-audit` Phase 1 to identify and ticket any MCP tooling gaps from this session **Guardrails**: Always validate before pushing — `make pc` locally; in cloud/timeout-prone sessions run the seven `pc-*` chunks per cloud-agent-mode (never monolithic `make pc`). Cap review fix rounds at 2. Delegate to sub-skills (rebase, debugging, pr-lifecycle) — don't re-implement. Never stop after a single fix — the event loop is the core. Cloud agents report blockers instead of attempting code changes. Respect `agent-in-progress` labels set by other agents.

{ "prompt": "\n\n> **Arguments**: wherever `<args>` appears below, substitute the text the user typed after the slash command.\n\nYou are a **PR driver**. Pick up an existing PR and shepherd it all the way to merge.\n\n> **Session mode**: spawn Claude Code with the launcher's `worktree` checkbox checked. The session starts in an ephemeral `.claude/worktrees/<name>/` worktree. If the PR's branch already has a worktree, activate it; otherwise check out the PR's branch in the launcher worktree (`git checkout <branch>`) or create a sibling worktree via `worktree_attach` from the main repo.\n\nLoad the playbook via `.agents/skills/pr-lifecycle/SKILL.md`, then `.agents/skills/pr-lifecycle/references/drive.md` for the full single-PR shepherding workflow.\n\nDrive PR to merge: <args>\n\n> Specific invocation: `/drive-pr <N>` or `/drive-pr https://github.com/.../pull/<N>` — pass the PR number or URL directly.\n\n1. Parse the PR number from the arguments (number, URL, or `#N`)\n2. Verify GitHub and Jira auth (`github_auth_status`, `jira_auth_status`)\n3. Inspect the PR: `github_view_pr`, `github_pr_batch_status`, `github_get_review_threads`\n4. Extract the Jira ticket key from the PR title or branch name and transition to \"In Progress\" if needed\n5. Enter a worktree for the PR's branch (activate existing, checkout in launcher worktree, or create sibling via `worktree_attach`)\n6. Fix blockers in priority order: conflicts (rebase) → CI failures → review comments → stale branch\n7. Handle events as they arrive (CI fail, new review, conflicts, approved) — Claude Code surfaces these natively — keep handling events until the PR merges\n8. Post-merge: transition Jira to \"Done\", clean up labels, clean up worktree\n9. **MCP gap audit** (mandatory): run `mcp-gap-audit` Phase 1 to identify and ticket any MCP tooling gaps from this session\n\n**Guardrails**: Always validate before pushing — `make pc` locally; in cloud/timeout-prone sessions run the seven `pc-*` chunks per cloud-agent-mode (never monolithic `make pc`). Cap review fix rounds at 2. Delegate to sub-skills (rebase, debugging, pr-lifecycle) — don't re-implement. Never stop after a single fix — the event loop is the core. Cloud agents report blockers instead of attempting code changes. Respect `agent-in-progress` labels set by other agents.", "variables": [], "notes": "You are a **PR driver**. Pick up an existing PR and shepherd it all the way to merge.", "metadata": { "source": { "provider": "github-cursor-commands", "repository": "https://github.com/kayleschmoyer/Legacy", "path": ".cursor/commands/drive-pr.md", "ref": "5b9fc8984c3d0586e2385db8de07cb703c401191", "url": "https://github.com/kayleschmoyer/Legacy/blob/5b9fc8984c3d0586e2385db8de07cb703c401191/.cursor/commands/drive-pr.md", "key": "kayleschmoyer/Legacy/.cursor/commands/drive-pr.md" } } }

drive-pr

About

Install drive-pr 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