Skip to content

Google ships user-scoped agent identity for MCP calls

Blog cover with the official CommandLyne logo: the headline Agents now act as the signed-in user in white type beside a three-step diagram of a signed-in user passing a delegated, scoped credential to an A2A agent with an attested SPIFFE identity, which then calls an MCP server or cloud API, on a dark navy grid.
CommandLyne TeamAI Product & Engineering

Nov 16, 20265 min read

An agent is about to open a customer record. It has a plan, a tool, and a permission check. The only open question left is whose name goes on the request.

For most teams shipping agents this year, the answer has been a shared service account with broad scopes. Google changed that answer on November 13, 2026, when three-legged OAuth (3LO) with the Agent Identity auth manager reached general availability for Agent2Agent (A2A) agents in Gemini Enterprise.

Primary source: Gemini Enterprise release notes (opens in a new tab).

What shipped

Three moving parts landed together.

PartWhat it doesWhere it runs
3-legged OAuth (3LO)The agent authenticates to a downstream system as the signed-in userA2A agents built with the Agent Development Kit (ADK) on Agent Runtime
Agent Identity auth managerValidates the consent callback and finalises the delegated credential in a Google-managed vaultGemini Enterprise
Agent Registry bindingsResolve which auth provider serves a given agent, MCP server, or APIConfiguration layer

The flow, in order:

  1. An A2A agent needs a downstream MCP server, a Google Cloud service, or a third-party API.
  2. If that resource requires consent, Gemini Enterprise prompts the signed-in user.
  3. The callback is validated, the credential is finalised by the Agent Identity auth manager, and the A2A conversation resumes on its own.
  4. The SDK carries the pieces: pip install --upgrade "google-adk[a2a,agent-identity,mcp]" with google-adk>=1.29.0.

The identity model underneath pairs a cryptographically attested SPIFFE agent identity with scoped, user-delegated 3LO credentials. Reference docs: Agent Identity (opens in a new tab) and 3-legged OAuth for A2A agents (opens in a new tab).

Why it matters: the credential is the control plane

Once an agent borrows the user's identity, three things become true at once. Scope shrinks to what that person can already reach. Every tool call carries a name. Revoking access is a single action on one account.

Google shipped the same idea at the browser layer. Chrome Enterprise now exposes GeminiActOnWebSettings (0 allowed, 1 blocked) alongside GeminiActOnWebAllowedForURLs and GeminiActOnWebBlockedForURLs, which scope exactly where an in-browser agent may act. When a URL matches both lists, the allowlist wins. GenAiDefaultSettings sets the organisation-wide default, and GeminiEnterpriseSettings binds a managed profile to a Gemini Enterprise project. Policy reference: Chrome Enterprise policy list (opens in a new tab).

Read the two releases together and the pattern is the same one, twice: an agent inherits a narrow, named, revocable grant, and the boundary is written down somewhere an admin can read.

            user signs in
                  |
                  v
     +---------------------------+
     |  signed-in user (3LO)     |
     +-------------+-------------+
                   | delegated, scoped
                   v
     +---------------------------+
     |  A2A agent on Agent       |   <- attested agent identity (SPIFFE)
     |  Runtime (ADK)            |
     +-------------+-------------+
                   | tool call + audit line
                   v
     +---------------------------+
     |  MCP server / Google      |
     |  Cloud service / API      |
     +---------------------------+

How CommandLyne handles this work

CommandLyne is the enterprise AI orchestration and governance platform where teams run Gemini, Claude, OpenAI, and partner models under their own keys (BYOK). It carries the same identity principle from the user seat through to the tool call.

Two ways to run a delegated action

StepAd-hoc agent scriptsGoverned routing in CommandLyne
IdentityShared service account with broad scopesSigned-in user, delegated and scoped
Tool accessEvery tool the runtime can reachPer-agent MCP tool policy
SpendDiscovered on the monthly invoiceToken-level cost visibility per task
ApprovalTrusted to run to completionPreview, Approve, Adjust, Cancel, Rollback
RecordLogs scattered across servicesFull audit trail with human sign-off
Model choiceOne model for every stepTask-complexity, role, latency, and cost routing

Where to start

  1. Inventory the identities your agents use today. List every service account, API key, and long-lived token sitting in an agent path.
  2. Move the first one to a delegated credential. Pick the agent that touches your most sensitive system and move it onto a user-scoped flow.
  3. Put checkpoints on the risky verbs. Writes, deletes, sends, and payments deserve a human sign-off step.
  4. Watch the cost line. Per-task token cost shows you which steps are worth a smaller model.

Try CommandLyne to run Gemini, Claude, and OpenAI agents under user-scoped credentials with per-agent tool policy, or book a demo with our team.

Try CommandLyne (opens in a new tab) | Book a demo (opens in a new tab)

See CommandLyne in action

Book a demo to see how CommandLyne brings your agents, models and workflows together under one governed workspace.