Skip to main content
POST
/
agent
/
{projectId}
/
job
Create agent job
curl --request POST \
  --url https://api.mintlify.com/v1/agent/{projectId}/job \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "branch": "<string>",
  "messages": [
    {
      "role": "system",
      "content": "<string>"
    }
  ],
  "asDraft": true,
  "model": "sonnet"
}
'
"<string>"
This endpoint creates an agent job based on provided messages and branch information. The job executes asynchronously and returns a streaming response with the execution details and results. If a branch doesn’t exist, the agent creates one. If files are edited successfully, a pull request is automatically created at the end of the job.

​
Rate limits

The agent API has the following limits:
  • 100 uses per Mintlify project per hour

​
Suggested usage

For best results, use the useChat hook from ai-sdk to send requests and handle responses.

Authorizations

​
Authorization
string
header
required

The Authorization header expects a Bearer token. Create an Admin API Key here.

Path Parameters

​
projectId
string
required

Your project ID. Can be copied from the API keys page in your dashboard.

Body

application/json
​
branch
string
required

The name of the Git branch that the agent should work on, will be automatically created if it doesn't exist

​
messages
object[]
required

A list of previous messages to provide to the agent.

​
asDraft
boolean
default:true

Control whether the pull request is created in draft or non-draft mode. When true, creates a draft pull request. When false, creates a regular (non-draft) pull request ready for review.

​
model
enum<string>
default:sonnet

The AI model to use for the agent job. Use sonnet for faster, cost-effective processing. Use opus for more capable, but slower processing.

Available options:
sonnet,
opus

Response

200 - text/plain

Agent job created successfully (streaming response). X-Session-Id Header is sent back in the response

Streaming response containing the agent job execution details and results.