Govern

Control what agents can access.

Full visibility into every action.

Credentials are scoped with per-endpoint controls to limit access. LLM interactions are mediated, and human approval flows can be enforced when needed. Governance is built directly into the runtime.

Start free trial Explore docs

Every agent gets its own identity
No more shared service accounts. No more hardcoded API keys.

01    agent_config:
02      auth:
03        api_key: "sk-prod-abc123"

Before

No control plane

01    github_integration:
02      - decision: ALLOW
03      - workspaces: [guildai/code-review]
04      - operations: [list_pull_requests, get_commit]
05      - resources:
06          - repos: [guildai/code-*]

Now

With Guild's control plane

Agents never see your API keys
Every LLM call is mediated through Guild. Enforce rate limits, track per-agent costs, apply content policies, and rotate credentials centrally.

Rate limiting

Per-agent, per-model spend thresholds.

Cost tracking

Token consumption attributed to specific agents.

Key rotation

Rotate credentials centrally, zero agent downtime.

Approval gates for high-risk actions

The agent proposes. A human approves. The control plane enforces the boundary. No agent executes a high-risk action without explicit approval.

Full session traces for every agent run.

Every LLM call, tool invocation, and decision is captured and visible in real time. When a SOC 2 auditor asks what happened, you can pull the full session trace.

Learn more about approvals and session traces

Explore docs

Track usage across agents, models, and teams.
Tokens consumed, cost per agent, cost per model, top users.

Active Agents

Monthly Spend

$0K

Incidents This Week

Team Members

Questions?
Start here.

Explore docs

Instead of agents calling LLM APIs directly with their own keys, Guild brokers every model call. This lets you enforce rate limits, track per-agent costs, apply content policies, and centrally rotate credentials — all without changing the agent code.

Each agent has its own identity with explicit permissions for each integration. You define which services an agent can access and which operations it can perform. Per-endpoint allow/deny. No shared service accounts.

For high-risk actions, the agent proposes, and a human approves before the action executes. The control plane enforces the boundary. No agent bypasses the approval gate.

Yes. Guild attributes token consumption and cost to specific agents, users, and models. Usage dashboards show exactly what your agents are doing and what they cost. No more surprise invoices.

SOC 2 certification is in progress. Guild provides full audit trails, scoped credentials, session tracing, and immutable logs. Everything an auditor needs to answer 'what happened and why.'

Agents ready?
Explore the full agent lifecycle.