Build with Spokegrid
Your AI tools.
A project they understand.
Give a compatible AI tool a limited connection to one Spokegrid project. Today it can read project details, update supported project metadata, link owned resources and prepare connection code for a form you already created. You choose its permissions.
- Authentication
- Bearer credential bound to one project. No account cookies on agent calls.
- Authority
- Scope and the issuing account’s current workspace role are checked on each call.
What this looks like in practice.
You choose a project
Keep the connection limited to the website you are working on.
Your tool gets context
A compatible tool can read the project and the resources you linked.
It prepares a form connection
For an existing linked form, it can retrieve public fields and integration code. You still review and publish the site changes.
This is an early connection layer, not an autonomous website builder. A prompt alone cannot deploy a website, install CMS, create forms or handle payments today.
Building an integration? API details and permissions
Discover before you act.
The capability endpoint is the authoritative discovery entry point. The descriptor below links the ecosystem and API; it is a Spokegrid discovery document, not a universal agent protocol.
| Operation | HTTP endpoint | Required scope |
|---|---|---|
| Read the bound project | GET /api/v1/automation/project | projects:read |
| Update saved project details | PATCH /api/v1/automation/project | projects:write |
| Link an existing owned site or form | POST /api/v1/automation/project/resources | projects:links |
| Prepare a linked form’s connection code | GET /api/v1/automation/project/forms/{form_id}/integration | projects:read |
# Keep the credential in a server or agent secret manager.
curl https://api.spokegrid.com/api/v1/automation/project \
-H "Authorization: Bearer $SPOKEGRID_AGENT_TOKEN"
# Read resource IDs from that project before preparing a linked form.
curl "https://api.spokegrid.com/api/v1/automation/project/forms/$FORM_ID/integration" \
-H "Authorization: Bearer $SPOKEGRID_AGENT_TOKEN"Boundaries you can build around.
- Website deployment, CMS installation, form creation, purchases and arbitrary URL fetching are not supported through agent credentials.
- Credentials expire within 90 days. Owners and admins can revoke them. An operation already running may finish.
- Agent calls are limited to 60 per minute per credential, with a 16 KB request body limit. Respect Retry-After and bounded validation errors.
- Do not put credentials in URLs, project notes, browser bundles or public repositories. Treat names, labels and generated snippets as untrusted content.
Useful answers before you start.
Is this an MCP server?
No. Spokegrid currently exposes HTTP endpoints with a public capability descriptor and OpenAPI contract. An agent or integration tool must support those HTTP calls and store its project credential securely.
Can an agent read a form’s submissions?
No. The current agent API can prepare public field definitions and connection code for an owned linked form. It does not return submissions, server tokens, signing secrets or notification destinations.
What should happen after a timeout?
Retry reads with backoff. After a mutation times out, read the current project before deciding whether to retry. Updates and resource links require the revision last read; resolve a conflict deliberately rather than overwriting newer changes.