Agentstudio APIs
Boomi provides two API solutions for managing AI agents:
- Agent Garden REST APIs build, package, and deploy agents within Boomi's managed platform.
- Agent Control Tower (ACT) GraphQL APIs register and manage agents from third-party providers.
Which API should you use?
| Use case | API | Learn more |
|---|---|---|
| Build, package, and deploy agents natively in Boomi | Agent Garden REST APIs | Full lifecycle automation |
| Catalog and manage agents from non-Boomi providers, such as Amazon Bedrock or custom LLMs | Agent Control Tower GraphQL APIs | Multi-provider catalog |
Agent Garden REST APIs
The Agent Garden REST APIs let you create, configure, package, and deploy AI agents in your Boomi account, from creation to live deployment in a runtime environment, without using the UI. The APIs follow REST conventions and use JSON for all request and response payloads.
Two API groups cover the full lifecycle:
- Design APIs build and configure agents. Refer to Design APIs overview.
- Execute APIs invoke agent sessions at runtime. Refer to Execute APIs overview.
Key features
- Agent management: Create agents by specifying objectives, tasks, guardrails, and custom configurations. Retrieve paginated lists of agents with optional filtering, or fetch the detailed configuration of a specific agent. Soft-deletion preserves records so you can restore an agent to
DRAFTstatus. - Packaging and deployment: Generate a snapshot of an agent's configuration, called a package, that captures the complete agent definition at the time of creation. You can optionally deploy to an environment in the same packaging request, or deploy the package later to make the agent available to others. Packages can be listed per agent or account-wide. You can delete a package if no active deployments reference it. Deployment records are retrievable per agent, per package, or account-wide, and you can remove deployments you no longer need.
- Agent execution: Invoke agent sessions at runtime through the Execute APIs.
Example: Multi-unit enterprise agent deployment
A financial services firm has 12 regional branches and a central support team. After building and testing a support-triage agent in Agent Studio, the firm rolls it out company-wide. The agent handles tier-1 ticket routing, policy lookups, and escalation recommendations. Each team has different needs:
- HQ support needs deep integrations, such as CRM connectors and custom logging, and has engineering resources.
- Regional branches lack engineering expertise and want to embed an agent widget on their websites with minimal setup.
- IT/DevOps wants all deployments managed through their CI/CD pipeline, with version control and automated rollouts across environments.
Agent Garden supports all three needs with the same agent, using three deployment paths.
Choose your deployment path
| Path | Best for | Effort |
|---|---|---|
| Execute API | Full control over the chat UI and integrations | Custom development |
| Design API and CI/CD | Versioned, automated rollouts across environments | Pipeline setup |
| EmbedKit CDN | Fast launch with no engineering resources | One-line embed |
Path 1: Execute API
Best for teams that need full control over the chat UI, custom integrations, or advanced analytics.
With this path, you can:
- Control the session lifecycle directly.
- Build your own chat interface in React, Vue, vanilla JS, or native mobile.
- Add custom logging, analytics, and CRM integration.
- Receive real-time responses through SSE streaming or JSON polling.
Choose this path if:
- You need a custom chat UI that matches your brand or product design.
- You're integrating with proprietary systems, such as CRM, ticketing, or a data warehouse.
- You need custom analytics or audit logging.
Example: The HQ support team embeds the agent in their proprietary CRM platform, threads agent responses into their case-management system, and logs interactions in a custom data warehouse.
Next step: Execute APIs documentation
Path 2: Design API and CI/CD pipeline
Best for organizations that manage deployments across development, staging, and production, and require version control and automated rollouts.
With this path, you can:
- Create and configure agents programmatically.
- Package agents as versioned snapshots.
- Automate deployment to multiple environments.
- Maintain a full audit trail with rollback capability.
- Manage everything in code instead of the UI.
Choose this path if:
- You manage multiple deployments, such as development, staging, and production.
- Your team uses Git for version control and CI/CD for releases.
- You need consistency, immutability, and rollback across environments.
- You want to treat agent configuration like code.
Example: DevOps keeps agent definitions in a Git repository. On merge to main, the CI/CD pipeline packages the agent, deploys it to staging, runs tests, and promotes it to production across all 12 regional environments.
Next step: Design APIs documentation
Path 3: EmbedKit CDN
Best for teams without engineering resources that need rapid deployment and don't need a custom UI.
With this path, you get:
- A prebuilt, themeable chat widget.
- A one-line embed code.
- Compatibility with any website through CORS support.
- No build tools, npm packages, or API integration.
Choose this path if:
- You need to launch in days, not weeks.
- Your team doesn't have frontend engineering capacity.
- A standard chat widget meets your UX needs and Boomi's built-in theming options are enough.
Example: Regional branch offices embed the agent on their public website with a single JavaScript snippet, with no custom integration.
Next step: EmbedKit CDN documentation
Combine paths
Many organizations combine paths:
- EmbedKit for public-facing, self-service experiences, such as branch websites and partner portals.
- Execute API for internal tools and custom integrations, such as CRM, support platforms, and internal dashboards.
- Design API to manage the agent across both, ensuring consistency.
Agent Control Tower GraphQL APIs
The ACT GraphQL APIs let you register, manage, and monitor AI agents from providers that Boomi doesn't natively support. Use them to catalog these providers and their agents alongside Boomi-supported ones, such as Boomi Garden, Salesforce Agentforce, Microsoft Copilot, Anthropic Managed Agents, and Amazon Bedrock, so you have a wider view of all your agents. For details, refer to Other providers.
The APIs expose GraphQL queries and mutations to retrieve, create, and update provider and agent data.
These GraphQL APIs support Custom Type provider accounts only. Running them on other provider types may produce incomplete or incorrect results because provider-specific fields are missing.
Why use the custom provider APIs
- Broader ecosystem support: Integrate and manage third-party and custom AI providers in ACT.
- Unified catalog: Centralize providers, agents, and their relationships with tasks, tools, instructions, and models in one place.
- Faster onboarding: Onboard new providers and agents directly through the APIs, without backend enablement.
- Consistency across UI and API: Providers and agents you create through the APIs appear automatically in the ACT interface.
- Scalability: Support new providers as they emerge, without waiting for native integrations.
- Auditable: Every creation, update, and deletion is logged for traceability and governance.
Learn more
- Agent Control Tower GraphQL APIs: Authentication and architecture.
- Operations: Available queries and mutations.
- Types: Object details.
Next steps
- Building and deploying Boomi-native agents? Start with the Agent Garden REST APIs overview.
- Embedding an agent quickly? Refer to EmbedKit CDN documentation.
- Building a custom integration? Refer to Execute APIs documentation.
- Managing deployments at scale? Refer to Design APIs documentation.
- Cataloging third-party agents? Refer to Agent Control Tower GraphQL APIs.
