Skip to main content

Alpha · v0.1.0 · Apache-2.0

The AI access layer for enterprises.

A self-hosted gateway that decides who may use which AI models and work apps, sets spending limits, keeps an audit record, and removes access when someone leaves.

Quick startGitHub
The Ambit console, Audit page: every model call one person made, with time, model, outcome and tokens.The Ambit console, Audit page: every model call one person made, with time, model, outcome and tokens.
The audit page of the Ambit console: every call one person made, with time, model, outcome and tokens.

Why Ambit

Without Ambit: Alice, Dan and Bob share one team API key and call models directly, an AI agent reaches work apps through a shared service account, Bob's key still works after he left, there is no record and the cost appears on next month's bill. With Ambit: everyone passes the Ambit gateway, which checks who they are, what they may use and their daily limit, records every call, and refuses Bob.Without Ambit: Alice, Dan and Bob share one team API key and call models directly, an AI agent reaches work apps through a shared service account, Bob's key still works after he left, there is no record and the cost appears on next month's bill. With Ambit: everyone passes the Ambit gateway, which checks who they are, what they may use and their daily limit, records every call, and refuses Bob.
Without AmbitWith Ambit
Each team shares one provider API key, copied into scripts and laptops.People sign in, or use a personal API key. Provider keys stay inside Ambit.
Anyone who has the key can use any model.Access rules based on your identity provider's groups decide which models and MCP servers each team may use.
AI agents connect to work apps on their own, often through one shared service account.Every MCP call passes one gateway. People connect their own app accounts, tools are checked, and personal data in results is masked.
The cost shows up on next month's bill.Daily limits per group and per person, in requests or in dollars.
Nobody can say who sent what to which model.Every call is recorded, refusals included. Model refusals do not yet show who made them (an agentgateway bug; see Known limitations).
Someone leaves, and their keys work until someone remembers to rotate them.The identity provider disables them, and their keys, sessions and app sign-ins stop at once.

Use cases

Architecture

Request path: clients call one Ambit gateway, built on agentgateway, which checks the caller and their access, asks the budget service, guard, token broker and router as needed, then forwards to AI models or MCP servers and sends audit records through the OpenTelemetry Collector to ClickHouse. Setup: your identity provider signs people in through Dex and reports leavers to the console by SCIM; admins use the console, which stores Ambit objects; ambit-controller turns them into gateway configuration.Request path: clients call one Ambit gateway, built on agentgateway, which checks the caller and their access, asks the budget service, guard, token broker and router as needed, then forwards to AI models or MCP servers and sends audit records through the OpenTelemetry Collector to ClickHouse. Setup: your identity provider signs people in through Dex and reports leavers to the console by SCIM; admins use the console, which stores Ambit objects; ambit-controller turns them into gateway configuration.

Blue is the path every request takes; grey is how admins and your identity provider set things up.

Built on

Ambit is built on these projects. Each keeps its own licence.

ProjectUse in AmbitLicence
agentgatewayLLM and MCP trafficApache-2.0
ToolHiveRuns MCP serversApache-2.0
DexSign-in and tokensApache-2.0
Gateway APIRouting objectsApache-2.0
ClickHouseAudit storageApache-2.0
OpenTelemetry CollectorAudit pipelineApache-2.0
ValkeyBudget countersBSD-3-Clause
envoyproxy/ratelimitBudget enforcementApache-2.0
MCP Go SDKPreset checks and test MCP serversApache-2.0 (older code MIT)