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.

multi-claude - Skill - OpenSmartRoute

Imported from thomasog/multi-claude (AGENTS.md). Install upstream with npx skills add thomasog/multi-claude. Copyright stays with the author.

AGENTS.md

Guidance for coding agents working on this repository. Humans: see CONTRIBUTING.md.

What this is

multi-claude runs several Claude Desktop accounts side by side. A profile is the official app started with its own --user-data-dir; icons are recolored from the installed logo; on Windows a hidden watcher tags the profile's windows with AppUserModelID MultiClaude.<name>. Read docs/how-it-works.md before changing behavior.

Scope

One job: several Claude Desktop accounts open side by side, each telling itself apart, working every time and putting nothing of the user at risk. Weigh every change against that.

In scope: what makes profiles open, stay apart and be told apart: launching, separate data, icons and colors, the taskbar button, keeping Claude in Chrome on the regular Claude, install and uninstall, and fixes when Claude changes.
Out of scope: what Claude already does itself or what serves another purpose: usage meters, account switching, syncing chats or settings between profiles, tray apps, auto-updaters.
Never: read or copy sign-ins or tokens, call Anthropic's services, patch or copy the app, kill Claude processes we didn't start, or touch the regular Claude's data.
Priorities: works correctly, then safe for the user, then stable, then fast. Prefer removing code to adding it. Workarounds for one Claude version live in one marked block and go once that version is gone.

Commands

powershell -ExecutionPolicy Bypass -File tests\windows.tests.ps1   # Windows tests
zsh tests/macos.sh                                                 # macOS tests
python3 -m unittest discover -s tests                              # recolor unit tests
bash tests/linux.sh                                                # Linux end-to-end tests
ruff check .                                                       # Python lint
shellcheck install.sh src/linux/multi-claude tests/linux.sh

Rules

The three CLIs (src/windows, src/macos, src/linux) expose the same commands, options, messages and exit codes (0 ok, 1 failure, 64 bad usage, and on Windows 3 for a terminal inside an app package). Change them together.
No runtime dependencies: PowerShell 5.1 + C# 5 on Windows, zsh and built-in tools on macOS, bash and python3 (standard library) on Linux.
Never commit Claude artwork, icons generated from it, or anything identifying a user.
install.ps1 and install.sh must keep their __VERSION__ and __SHA256__ placeholders; the release workflow fills them in.
Keep the README short. Details belong in docs/.

Install multi-claude 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.

multi-claude

About

AGENTS.md

What this is

Scope

Commands

Rules

Install multi-claude 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