Skip to main content

​
HTTP

​
GET /health

Returns server status. 
{ "ok": true }

​
GET /providers

Lists available providers and their capabilities. 
{
  "providers": [
    {
      "id": "claude",
      "label": "Claude",
      "defaultModel": "claude-sonnet-4-6",
      "models": ["claude-sonnet-4-6", "claude-opus-4-7", "claude-haiku-4-5"],
      "configured": true
    }
  ]
}

​
POST /chat

Send chat messages to a provider. Streams SSE by default. Request body:
FieldTypeRequiredDescription
providerstringyescodex or claude
modelstringyesProvider model ID
messagesarrayyes{ role, content } chat messages
systemPromptstringnoSystem prompt
streambooleannoDefaults to true
Example: 
curl -N -X POST http://localhost:4096/chat \
  -H "Content-Type: application/json" \
  -d '{
    "provider": "codex",
    "model": "gpt-5.4-mini",
    "messages": [{ "role": "user", "content": "Explain this module." }]
  }'

​
POST /completions

Return a one-shot completion for an editor buffer.
FieldTypeRequiredDescription
prefixstringyesText before the cursor
suffixstringyesText after the cursor
languagestringyesFile language
filepathstringyesCurrent file path
maxTokensnumbernoCompletion budget
temperaturenumbernoSampling temperature
providerstringnoProvider override
modelstringnoModel override

​
POST /inline-edit

Return replacement text for a selected range.
FieldTypeRequiredDescription
filestringyesFile path
rangeobjectyesstartLine, startColumn, endLine, endColumn
promptstringyesEdit instruction
textstringyesSelected text
repoRootstringnoRepository root
providerstringnoProvider override
modelstringnoModel override

### `GET /sessions`

Lists all active sessions.

```json
[
  {
    "id": "abc-123",
    "repoRoot": "/repo",
    "provider": "claude",
    "model": "claude-sonnet-4-6",
    "createdAt": "2026-04-26T10:00:00.000Z"
  }
]

​
GET /sessions/:id

Get a single session by ID.

​
DELETE /sessions/:id

Delete a session and free its resources.

​
SSE Events

When streaming from POST /chat, the server emits data: lines containing one of these event objects:
EventDescription
startProvider and model selected
text_deltaIncremental text chunk from the model
doneFinal text and optional usage
errorError during processing
Example stream: 
data: {"type":"start","provider":"codex","model":"gpt-5.4-mini"}
data: {"type":"text_delta","text":"This module"}
data: {"type":"done","provider":"codex","model":"gpt-5.4-mini","text":"This module..."}

​
WebSocket

Connect to: 
ws://localhost:4096/ws
The server sends a ready message first: 
{ "type": "ready", "protocolVersion": 1 }

​
Session Messages

| Client message | Purpose | |---|---|---| | session.create | Create a git-backed agent session | | session.open | Reopen an existing session | | chat.turn | Run an agent turn in the session worktree | | run.cancel | Cancel an active session run | | changes.status | Read changed files | | changes.accept_all | Apply all agent changes to the repo | | changes.discard_all | Drop all agent changes | | changes.accept_file | Apply one changed file | | changes.discard_file | Drop one changed file |

​
One-Shot Messages

Client messagePurpose
chatStream a stateless chat request
completionStream a stateless completion request
inline.editReturn replacement text for an edit
cancelCancel a one-shot request by ID
pingHealth check

​
Server Messages

Server messagePurpose
session.createdSession metadata
assistant.deltaAgent text chunk
assistant.doneAgent turn complete
inline.resultInline edit replacement
changes.updatedChanged files and refs
run.statussyncing, running, checking, checkpointing, or done
chat_eventStateless chat or completion stream event
cancelledRequest cancelled
pongPing response
errorError message