Develofly Autopilot
Home › API

Autopilot API and AI assistants

Start bots, read their results and manage schedules from your own scripts, or let an AI assistant do it for you. The same keys, scopes, rate limits and audit log cover the local API, the cloud API and MCP.

Download Develofly Autopilot MCP setup

Overview

Autopilot can be controlled by other apps and by AI assistants (Claude Code, Cursor, GitHub Copilot, Codex, LM Studio) through a small REST API and an MCP server. Requests and answers are JSON. Every request sends Authorization: Bearer DFAPI_… (or DFCAPI_… for the cloud API). Keys in the URL are ignored.

Command line: develofly-autopilot api-key --name ci --scope run (also --list, --revoke ID). develofly-autopilot api runs the API without the window, develofly-autopilot mcp runs the MCP server over stdio, and develofly-autopilot mcp --http serves MCP at /mcp on the API port.

Turn it on

The API is off until you turn it on. In Autopilot open Settings → API and AI assistants and tick Let other apps and AI assistants control Autopilot on this computer. The API then answers at http://127.0.0.1:8790 (the port can be changed), and only programs on the same computer can reach it.

Allow other computers to connect (remote access) is a separate switch and is also off until you turn it on. Turning it on asks first: other computers can then reach Autopilot, anyone with a key can use that key's permissions, and the connection is not encrypted. Only use it on a network you trust, or behind an HTTPS tunnel. With remote access off, requests from other computers are refused even if they reach the port, and only 127.0.0.1/localhost Host headers are accepted. Requests that carry a browser Origin header are refused, so web pages can't call the API.

Keys and scopes

Create a key with New key…. Autopilot shows the key once, so copy it then. Only a SHA-256 fingerprint of it is stored. Revoke stops a key immediately. Each key is limited to 60 requests a minute (per key and per caller address); over that the API answers 429. Every call is written to the audit log in Settings (time, key id, method, route template, status and caller address — no ids, no query strings, no bodies).

ScopeWhat it allows
read-onlyList bots, read runs, results, logs, screenshots and schedules.
runEverything in read-only, plus start and stop runs and trigger flows.
editEverything in run, plus create and change bots and their steps, schedules, and build a bot from a description.
adminEverything, including creating and revoking keys and reading the audit log.

Endpoints

Local paths start with /api/v1 on http://127.0.0.1:8790 (so a full URL looks like http://127.0.0.1:8790/api/v1/projects).

RequestScopeWhat it does
GET /healthnoneIs the API up.
GET /capabilitiesread-onlyWhat this API can do here.
GET /projects, GET /botsread-onlyProjects and bots in the workspace (opaque ids).
GET /projects/{id}read-onlyOne bot and its steps. Locked bots show no steps.
POST /projectseditCreate a bot from a list of steps and inputs.
PUT /projects/{id}editRename a bot or change its steps.
POST /projects/{id}/runsrunStart a run. Send {"inputs": {...}} for one row or {"rows": [...]} for many. Answers 202 with the run id.
GET /runs, GET /runs/{id}read-onlyRun status: running, stopping, ok, completed (some rows failed), stopped or error.
POST /runs/{id}/stoprunStop after the current step.
GET /runs/{id}/resultsread-onlyResult rows. Columns that look like passwords or keys are hidden.
GET /runs/{id}/logsread-onlyThe run log.
GET /runs/{id}/screenshots, GET /runs/{id}/screenshots/{name}read-onlyScreenshots the bot saved (base64 PNG).
GET, POST /schedules · PUT, DELETE /schedules/{id}read-only / editDaily, weekly, every-N-minutes and health-check schedules. Your plan's schedule limit applies.
POST /build-boteditDescribe a bot in plain English and get proposed steps back. This uses your plan's AI allowance, or your own AI key if you set one.
POST /flows/{id}/triggerrunStarts a flow from a webhook or chat. Answers 501 until flows are part of your Autopilot.
GET /keys, POST /keys, DELETE /keys/{id}, GET /auditadminKey management and the audit log.

Runs started through the API run unattended. A step that needs a person stops that row instead of waiting.

Examples

List your bots

curl

curl -s -H "Authorization: Bearer $DEVELOFLY_API_KEY" \
     http://127.0.0.1:8790/api/v1/projects

Python

import os, requests

API = "http://127.0.0.1:8790/api/v1"
H = {"Authorization": f"Bearer {os.environ['DEVELOFLY_API_KEY']}"}

for bot in requests.get(f"{API}/projects", headers=H, timeout=30).json()["projects"]:
    print(bot["id"], bot["name"])

JavaScript

const API = "http://127.0.0.1:8790/api/v1";
const H = { Authorization: `Bearer ${process.env.DEVELOFLY_API_KEY}` };

const { projects } = await (await fetch(`${API}/projects`, { headers: H })).json();
for (const bot of projects) console.log(bot.id, bot.name);

Run a bot and read the results

curl

# start a run with inputs (needs a key with the run scope)
curl -s -X POST -H "Authorization: Bearer $DEVELOFLY_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"inputs": {"city": "Austin"}}' \
     http://127.0.0.1:8790/api/v1/projects/$BOT_ID/runs

# check it, then read the results
curl -s -H "Authorization: Bearer $DEVELOFLY_API_KEY" http://127.0.0.1:8790/api/v1/runs/$RUN_ID
curl -s -H "Authorization: Bearer $DEVELOFLY_API_KEY" http://127.0.0.1:8790/api/v1/runs/$RUN_ID/results

# stop it
curl -s -X POST -H "Authorization: Bearer $DEVELOFLY_API_KEY" http://127.0.0.1:8790/api/v1/runs/$RUN_ID/stop

Python

import os, time, requests

API = "http://127.0.0.1:8790/api/v1"
H = {"Authorization": f"Bearer {os.environ['DEVELOFLY_API_KEY']}"}

run = requests.post(f"{API}/projects/{bot_id}/runs", json={"inputs": {"city": "Austin"}},
                    headers=H, timeout=30).json()["run"]
while True:
    status = requests.get(f"{API}/runs/{run['id']}", headers=H, timeout=30).json()["run"]["status"]
    if status not in ("running", "stopping"):
        break
    time.sleep(5)
rows = requests.get(f"{API}/runs/{run['id']}/results", headers=H, timeout=30).json()["results"]
print(status, len(rows), "rows")

JavaScript

const API = "http://127.0.0.1:8790/api/v1";
const H = { Authorization: `Bearer ${process.env.DEVELOFLY_API_KEY}`, "Content-Type": "application/json" };

const { run } = await (await fetch(`${API}/projects/${botId}/runs`, {
  method: "POST", headers: H, body: JSON.stringify({ inputs: { city: "Austin" } }),
})).json();

let status = "running";
while (status === "running" || status === "stopping") {
  await new Promise((r) => setTimeout(r, 5000));
  ({ run: { status } } = await (await fetch(`${API}/runs/${run.id}`, { headers: H })).json());
}
const { results } = await (await fetch(`${API}/runs/${run.id}/results`, { headers: H })).json();
console.log(status, results.length, "rows");

Schedule a bot

curl

curl -s -X POST -H "Authorization: Bearer $DEVELOFLY_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"project_id": "'$BOT_ID'", "kind": "weekly", "at": "07:30", "days": [0, 2, 4]}' \
     http://127.0.0.1:8790/api/v1/schedules

Python

requests.post(f"{API}/schedules", headers=H, timeout=30,
              json={"project_id": bot_id, "kind": "daily", "at": "08:00"})

JavaScript

await fetch(`${API}/schedules`, { method: "POST", headers: H,
  body: JSON.stringify({ project_id: botId, kind: "interval", every_minutes: 60 }) });

Build a bot from a description

curl

curl -s -X POST -H "Authorization: Bearer $DEVELOFLY_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"description": "Open example.com/products, collect each product name and price", "save": true}' \
     http://127.0.0.1:8790/api/v1/build-bot

Python

proposal = requests.post(f"{API}/build-bot", headers=H, timeout=180, json={
    "description": "Log in to my supplier portal and download this week's invoices",
    "url": "https://supplier.example.com/login",
}).json()["proposal"]
print(proposal["explanation"])
for step in proposal["steps"]:
    print("-", step["type"])

JavaScript

const { proposal } = await (await fetch(`${API}/build-bot`, { method: "POST", headers: H,
  body: JSON.stringify({ description: "Fill the contact form on example.com for each row", columns: ["name", "email"] }) })).json();
console.log(proposal.explanation);

Cloud runs

Bots you run in the cloud (Pro and Agency plans with cloud hosting) have the same API at https://bot.develofly.com/api/v1/public with separate cloud keys (DFCAPI_…) that you create in Autopilot. You can list cloud bots, start and stop runs, read results and logs, and change schedules. Cloud bots are encrypted, so creating or editing steps, creating or deleting schedules, building a bot and flow triggers answer 501; those changes are made in Autopilot and uploaded again. Screenshots return an empty list. Cloud keys stop working if the subscription or cloud hosting ends. The default rate limit is 60 requests a minute.

curl

curl -s -H "Authorization: Bearer $DEVELOFLY_CLOUD_KEY" https://bot.develofly.com/api/v1/public/projects
curl -s -X POST -H "Authorization: Bearer $DEVELOFLY_CLOUD_KEY" https://bot.develofly.com/api/v1/public/projects/$JOB_ID/runs

Python

CLOUD = "https://bot.develofly.com/api/v1/public"
CH = {"Authorization": f"Bearer {os.environ['DEVELOFLY_CLOUD_KEY']}"}
run = requests.post(f"{CLOUD}/projects/{job_id}/runs", headers=CH, timeout=30).json()["run"]
print(requests.get(f"{CLOUD}/runs/{run['id']}/logs", headers=CH, timeout=30).json()["logs"])

JavaScript

const CLOUD = "https://bot.develofly.com/api/v1/public";
const CH = { Authorization: `Bearer ${process.env.DEVELOFLY_CLOUD_KEY}` };
const { runs } = await (await fetch(`${CLOUD}/runs`, { headers: CH })).json();

AI assistants (MCP)

Autopilot includes an MCP server, so AI assistants can list and run your bots, read results, manage schedules and build a new bot from a description. Every tool call goes through the same API, so the key's scope, rate limit and audit log apply. In Settings pick your assistant and press Copy setup, or paste one of these and put in your key (the $DEVELOFLY_API_KEY value). A read-only key is a good start; give the assistant a run or edit key only if you want it to start or change bots.

stdio: develofly-autopilot mcp with DEVELOFLY_API_KEY in the environment (newline-delimited JSON-RPC 2.0). HTTP: POST http://127.0.0.1:8790/mcp (stateless Streamable HTTP, JSON responses; GET answers 405). Protocol versions 2025-06-18, 2025-03-26 and 2024-11-05.

Tools: list_bots, get_bot, run_bot, run_status, stop_run, run_results, run_logs, list_screenshots, get_screenshot, list_schedules, create_schedule, update_schedule, delete_schedule, create_bot, update_bot_steps, build_bot_from_description, trigger_flow, capabilities.

Claude Code

claude mcp add develofly --env DEVELOFLY_API_KEY=$DEVELOFLY_API_KEY -- develofly-autopilot mcp

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "develofly": {
      "command": "develofly-autopilot",
      "args": ["mcp"],
      "env": { "DEVELOFLY_API_KEY": "$DEVELOFLY_API_KEY" }
    }
  }
}

GitHub Copilot in VS Code (.vscode/mcp.json)

{
  "servers": {
    "develofly": {
      "type": "stdio",
      "command": "develofly-autopilot",
      "args": ["mcp"],
      "env": { "DEVELOFLY_API_KEY": "$DEVELOFLY_API_KEY" }
    }
  }
}

Codex (~/.codex/config.toml)

[mcp_servers.develofly]
command = "develofly-autopilot"
args = ["mcp"]
env = { DEVELOFLY_API_KEY = "$DEVELOFLY_API_KEY" }

LM Studio (mcp.json)

{
  "mcpServers": {
    "develofly": {
      "command": "develofly-autopilot",
      "args": ["mcp"],
      "env": { "DEVELOFLY_API_KEY": "$DEVELOFLY_API_KEY" }
    }
  }
}

Any assistant that connects over HTTP (Autopilot open, API turned on)

{
  "mcpServers": {
    "develofly": {
      "url": "http://127.0.0.1:8790/mcp",
      "headers": { "Authorization": "Bearer $DEVELOFLY_API_KEY" }
    }
  }
}