The Control Plane for AI Agents

Every engineer is shipping agents.No one is governing them.

See every agent. Every action. Every dollar.

Token Usage-2% from yesterday ↓
Tokens: 30 day317M
7d30d90d
Agent Spend+12% last 30d ↑
Monthly total$802.50
PR Reviewer$438.50
Issue Triage$206.20
Deploy Notifier$157.80
LLM GatewayLive
Requests today265,019
Anthropicclaude-sonnet-4
38.4/50
OpenAIgpt-4o
24.1/40
Googlegemini-2.5-pro
12.7/25
Credentials8 Connected
GitHubDanny K: 5/12/26
Connected
SlackAlex M: 5/08/26
Connected
LinearSteph T: 4/28/26
Connected
JirraPat K: 4/03/26
Connected
ZendeskJames E: 3/25/26
Connected
Triggers4 configured
Jiraissue.created
Webhook
GitHubpull_request.opened
Webhook
Cron*/15 * * * *
Schedule
PagerDutyincident.triggered
Webhook
Permissions
5 allow1 deny
GETgetPullRequest
GETlistBranches
POSTcreateReview
GETlistCommits
DELETEdeleteRepository
POSTcreateReview
Drag

One Control Plane to Rule Them All

Token Usage-2% from yesterday ↓
Tokens Today20.7M
Sessions4,291
Tokens317M
Agent Spend+12% last 30d ↑
Monthly Total$802.50
PR Reviewer$438.50
Issue Triage$206.20
Deploy Notifier$157.80
LLM GatewayLive
Requests Today265,019
Key Rotation
1h 42m
Anthropicclaude-sonnet-4
38.4/50
OpenAIgpt-4o
24.1/40
Googlegemini-2.5-pro
12.7/25
Credentials8 Connected
GitHubDanny K: 3/12/26
Connected
SlackAlex M: 3/08/26
Connected
TriggersAll
jirra / pull requestAction: All
Active
linear / issueAction: Opened
Active
Permissions
12 Allow3 Deny
GETgetPullRequest
DELETElistCommits
GETcreateReview

You can finally answer the 3am question.

One place to see every agent, govern what it can access, and track what it costs.

THE DISCONNECT

Sound familiar?
Every engineering org shipping agents is living this.

Credentials shared across agents
No one knows where they’re scoped
Your AI bill doubled last quarter
No one can tell you which agents drove it
An agent went off-script in production
No one knows when it started
Four teams building the same PR review agent
None talking to each other
Security asking “what agents have access to what”
No one has an answer

These aren't tooling problems. They're governance gaps.

See how Guild fixes this
The category

Agents need a control plane
Not another framework

Guild is the control plane for AI agents
A control plane is the layer that manages how agents deploy, what they access, and what happens when something goes wrong.

Your agents run through it, not alongside it.

That's the difference between tools that watch agents and a platform that governs them.
BYOKUse your own LLM keys from any provider
Mediated accessAgents never see your API keys directly
Usage trackingToken consumption per agent, per model
No vendor lock-inSwitch providers without changing agents

Model-agnostic. Your stack. Your keys.
Start with Guild's agents and integrations, then extend them with your own providers, and custom workflows.

Governance, visibility, and control at scale.Explore Enterprise
Code-first. TypeScript SDK. Full control.Explore Developers

Bring your own models

Guild works across models and vendors. Teams can use OpenAI, Anthropic, Google. Guild keeps you flexible, so agents aren't tied to a single model vendor or setup.
OpenAIGPT4-0, o1, o3
OpenAI
AnthropicClaude 4, Sonnet, Haiku
Anthropic
PerplexitySonar, Online
Perplexity
GoogleGemini
Google

Works with your credentials
Connect the systems your team already uses, like GitHub, Slack, Linear, Jira, and more. Connect to your docs and your internal tools. Use Guild to automate your existing workflows with the systems and tools you already have.

Jira
Jira
BitBucket
BitBucket
Google Docs
Google Docs
Linear
Linear
TestRail
TestRail
Azure
Azure
Slack
Slack
Cypress
Cypress
Confluence
Confluence
GitHub
GitHub
New Relic
New Relic
Figma
Figma
Zendesk
Zendesk
Google Compute
Google Compute
GCP
GCP
Notion
Notion
Turbopuffer
Turbopuffer
Docker
Docker
Azure DevOps
Azure DevOps
AWS
AWS
Google Logging
Google Logging

Integration endpoints

CLI
$ guild credentials endpoint list acme~github

OPERATION        METHOD  PATH
repos_list       GET    /orgs/{org}/repos
repos_get        GET    /repos/{owner}/{repo}
pulls_list       GET    /repos/{owner}/{repo}/pulls
pulls_create     POST   /repos/{owner}/{repo}/pulls
issues_list      GET    /repos/{owner}/{repo}/issues
issues_create    POST   /repos/{owner}/{repo}/issues
gists_list       GET    /gists
gists_create     POST   /gists

Showing 8 of 42 endpoints

Bring your own integration

Connect anything your agents need — custom APIs and REST endpoints. Guild fits into your stack seamlessly.
MCP server support
MCP server support
Custom API connections
Custom API connections
REST endpoint integrations
REST endpoint integrations
More connectors shipping continuously
More connectors shipping continuously
The lifecycle

One platform. Four pillars.
Every stage covered. Build, deploy, govern, and share agents — in one control plane.

1

Build

Any surface you like — SDK, CLI, or UI. Start with production-ready starters from Agent Hub, or scaffold from scratch.

Open-source TypeScript SDK
CLI: guild init / test / deploy / publish
Agent Hub with 40+ starter agents
MCP + custom REST endpoints
2

Deploy

Any framework, any model. Guild handles the runtime so you don’t. Event-driven by default — agents run when the world changes.

Managed runtime
Event triggers
Version control + rollback
Model-agnostic
3

Govern

This is where most platforms fall short. Guild treats governance as architecture, not a feature.

Scoped credentials
Sandboxed execution
Immutable audit trail
Policy engine + approval gates
4

Share

Stop rebuilding the same agent four times across four teams. Guild treats agents like source code — versioned, forkable, discoverable.

Shared registry
Public + private agent hub
Multiplayer by default
Cross-team visibility

Teams closing the governance gap
See how engineering teams use Guild across support, triage, code review, and internal tools.

Turo

Our early work with Guild.ai has been instrumental as we expand how we use AI agents at Turo.

Paul Velez, SVP, Engineering
Paul Velez, SVP, EngineeringTuro
Sovrn

Guild.ai gives us the visibility and governance layer we were missing: a single place to manage how agents interact with our systems, what they can access, and what they cost.

Jud Valeski, CTO
Jud Valeski, CTOSovrn
WorkWhile

With Guild.ai we now have a structured way to deploy and govern these agents without having to build the management layer ourselves.

Matt Blair, Head of Engineering
Matt Blair, Head of EngineeringWorkWhile

The hard questions
What engineering leaders actually ask

Those are monitoring tools — they tell you what happened. Guild is a control plane — it governs what can happen. Monitoring watches agents after the fact. A control plane scopes credentials, enforces policies, and audits every action before anything goes wrong. You run both; they do different jobs.

No. Guild is model-agnostic (Claude, GPT, Gemini, Llama, custom), framework-agnostic (LangChain, CrewAI, Mastra, or roll your own), and vendor-neutral by design. Bring your own integrations via MCP or custom REST. The only thing you commit to is the governance layer.

SOC 2 certification is in progress. Guild ships with sandboxed execution, tenant isolation, scoped credentials (deny-by-default), and immutable audit trails — the architecture that SOC 2 certification formalizes.

No. Bring existing agents as-is. Guild wraps them with the control plane — scoped credentials, audit trail, policy gates — without forcing a rewrite. Your LangChain, CrewAI, or custom TypeScript agents run on Guild without code changes.

Free to start. Starter tier at $199/mo for 1,500 agent automations and 500k tokens. Enterprise is custom (SSO, SCIM, RBAC, audit logs, extended retention, SLAs). Full detail at guild.ai/pricing.

VPC and self-hosted deployment are on the roadmap. Today: Guild runs as a managed cloud service with tenant isolation. Enterprise contracts include extended deployment options — talk to sales if that's a hard requirement.

One control plane.
The complete agent lifecycle.
Get a working agent in under 30 minutes.
No credit card required.