> For the complete documentation index, see [llms.txt](https://docs.visdom.virtuslab.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.visdom.virtuslab.com/visdom-ai-gateway/readme.md). # Visdom AI Gateway One governed door to every model and every tool your organisation uses. AI use spreads faster than its governance. Every IDE assistant, agent and internal application brings its own provider key, its own idea of which model is allowed, and no shared record of what it cost or what it sent. The Visdom AI Gateway puts one door in front of all of it: every model call and every tool call goes through the same address, with an owner, a budget, a policy and a record. It has two faces: * **LLM Gateway** – one OpenAI- and Anthropic-compatible endpoint in front of every model provider and every model served inside the network. Clients see model aliases; which provider, region and contract serves an alias is decided centrally. * **MCP Gateway** – one MCP endpoint in front of every approved tool server, with a registry, per-server credentials the user never sees, and an admin UI. Both run on the organisation's own infrastructure, work without internet access, and send no telemetry. Models run on the organisation's own provider contracts; Visdom never resells model usage. ## What Visdom adds The two data planes are mature open source: [LiteLLM](https://github.com/BerriAI/litellm) for models and [IBM MCP Context Forge](https://github.com/IBM/mcp-context-forge) for tools. What the market does not sell as one product is the layer around them, and that is the Visdom part: * **Identity and budgets from the Control Plane** – the same people, teams and projects as every other Visdom product. * **A policy decision before every call** – Policy Gate: allowed or not, and why, deny by default. * **A signed record and a cost for every call** – AI Tracing: one evidence chain and one cost ledger for the IDE, background agents and applications alike. ## Available today The details, and how each item was verified, are on [Available today](/visdom-ai-gateway/available-today.md). * **A tested reference deployment of both gateways** – LiteLLM 1.100.1 and MCP Context Forge 1.0.10 on PostgreSQL, in the `visdom-ai-gateway` repository, with a smoke test that needs no model provider. * **Virtual keys, team budgets and model allow-lists** – an alert at a soft budget (for example 80%), a hard stop at the limit, spend attributed to the team, models outside a team's list refused. * **Both wire formats on one key** – OpenAI Chat Completions and Anthropic Messages. * **An MCP gateway** that registers tool servers, discovers their tools, groups them into virtual servers and serves them over MCP Streamable HTTP, with authentication required. * **Visdom pieces already running around it** – the AI Tracing proxy (Anthropic Messages and OpenAI Responses with per-user keys, secret blocking and a per-call cost ledger), per-agent MCP allow-lists and human approvals in the Orchestrator, and Sandcat to route workstation tools only to the gateway. ## Roadmap The details are on [Roadmap](/visdom-ai-gateway/roadmap.md). * **Next:** keys, teams and budgets issued by the Control Plane; every gateway call written to AI Tracing's signed chain; Policy Gate v1 in both gateways; one kill switch for a model, provider, MCP server or tool; PII masking; SIEM export; a Helm chart. * **Later:** the organisation hierarchy with inherited budgets; per-user OAuth to MCP tools and tool-level access control enforced by the gateway; content classifiers for prompt injection and unsafe output; throttling or rerouting to a cheaper model at a budget threshold; the current MCP specification; signed offline update bundles; high availability. {% hint style="info" %} This section reflects the `visdom-ai-gateway` repository and the upstream versions named above, read on **14 September 2026**. Roadmap items are direction, not commitments; order and scope can change. {% endhint %} --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.visdom.virtuslab.com/visdom-ai-gateway/readme.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.