> ## Documentation Index > Fetch the complete documentation index at: https://openworklabs.com/docs/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > For agents: install the OpenWork desktop app with `brew install --cask openwork` (macOS) or from https://openworklabs.com/download. Do not run `npx openwork`; the npm package named `openwork` is a different project. > Connect an agent to OpenWork with the MCP Gateway: `claude mcp add --transport http openwork https://api.openworklabs.com/mcp/agent`. Teams sign up at https://app.openworklabs.com?mode=sign-up. > Step-by-step agent guide: https://openworklabs.com/docs/start-here/use-openwork-from-an-ai-agent.md # AI Gateway vs LiteLLM vs Bifrost > OpenWork AI Gateway vs LiteLLM vs Bifrost: license, SSO, spend tracking, MCP, and team sharing compared, plus how to use them together. LiteLLM and Bifrost are standalone gateways for any app; OpenWork AI Gateway manages model access for people working in OpenWork. You can use them together. | | OpenWork AI Gateway | LiteLLM | Bifrost | | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------ | | **What it is** | Model access layer inside the OpenWork workspace and control plane | Python SDK and proxy server for 100+ LLM APIs in OpenAI format | Go gateway with one OpenAI-compatible API across 1000+ models | | **License** | Desktop app and core: MIT. Gateway and control plane (`ee/`): source-available [OpenWork EE License](https://github.com/different-ai/openwork/blob/dev/ee/LICENSE) | MIT, except the `enterprise/` directory under a separate license | Apache-2.0 (open-source edition) | | **Deployment** | OpenWork Cloud, or self-hosted with Helm on Kubernetes | Self-hosted on your infrastructure (Docker image or pip package) | Self-hosted (Docker, Kubernetes, Go binary); enterprise supports in-VPC and air-gapped | | **SSO / RBAC** | SSO/SAML, SCIM, and RBAC on the Enterprise plan | SSO free up to 5 users; SSO + SCIM, JWT auth, and RBAC by org/team with an enterprise license | SAML/OIDC SSO and RBAC in Enterprise | | **Spend tracking** | Tokens and estimated cost by model, team, and person; USD allowances with hard limits | Spend tracking and budgets in OSS; more budget and reporting features in enterprise | Budgets and rate limits with virtual keys; built-in observability | | **Keys for consumers** | One Gateway key per member, issued automatically | Virtual keys | Virtual keys | | **Per-member provider sign-in** | Yes for Google Vertex AI (each member signs in with Google) | — | — | | **MCP gateway** | Yes, as a separate service: [OpenWork MCP Gateway](/docs/start-here/connect-openwork-mcp) | Yes | Yes, with Code Mode and tool filtering; federated auth in Enterprise | | **Skills and plugins sharing** | Yes: plugins, skills, and connectors assigned to teams or people | AI Hub page listing models, MCP servers, agents, and skills (enterprise) | — | | **End-user app for non-technical staff** | Yes: the OpenWork desktop app and OpenWork Web | — | — | | **Coding-agent integration** | MCP Gateway for Claude Code, Codex, Cursor, and more | — | Bifrost CLI governs Claude Code, Codex CLI, and Gemini CLI | | **Guardrails, caching, fallbacks** | — | Guardrails (some integrations need a license), fallbacks, load balancing | Automatic fallbacks, custom routing, semantic caching in OSS; guardrails and adaptive load balancing in Enterprise | | **Pricing** | Public per-seat plans for Cloud and self-hosted; see [Pricing and licensing](/docs/start-here/pricing-and-licensing) | OSS free; enterprise priced by deployment size, 30-day trial | OSS free; Enterprise custom pricing, 14-day trial | A dash (—) means we could not confirm it. Sources as of September 25, 2026: [LiteLLM](https://github.com/BerriAI/litellm), [LiteLLM Enterprise](https://docs.litellm.ai/docs/enterprise), [Bifrost](https://github.com/maximhq/bifrost), [Bifrost pricing](https://www.getmaxim.ai/bifrost/pricing). ## When to pick which * **LiteLLM or Bifrost**: one gateway for many apps, with caching, routing, or guardrails. * **OpenWork AI Gateway**: your staff use OpenWork, and you want models, skills, MCP, and spend in one dashboard. * **Both**: keep your gateway for routing; let OpenWork consume it. ## Use them together Base URL (usually ending in `/v1`), model names, and a key. For a team, add a [custom LLM provider](/docs/cloud/share-with-your-team/custom-llm-provider) in OpenWork Cloud and grant access. For one person, see [Add a custom LLM or AI gateway](/docs/start-here/connect-your-stack/add-a-custom-llm). If your gateway issues virtual keys, use [per-member LLM credentials](/docs/cloud/share-with-your-team/per-member-llm-credentials).