Cursor connector for AI agents

Tools your coding agent reaches for on Cursor, scoped per user.

CALL ANY TOOL

24 tools for Cursor Cloud Agents: create coding agents and send follow-up prompts, track runs and token usage, list artifacts, and manage saved cloud agent environments and their builds.

cursor_agent_create

Create agent

Create a Cursor Cloud Agent and start its first run on a prompt, optionally against a GitHub repository, branch, or pull request. Returns the new agent (id, name, status, url, repos) and its first run (id, status CREATING or RUNNING). Use this to start new work. Use cursor_run_create to send a follow-up prompt to an existing agent. Poll cursor_run_get with the returned agent and run IDs to see when the run finishes.

Parameters

Name

Type

Required

Description

prompt

string

Required

Task instruction for the agent, in plain language. This becomes the prompt of the agent's first run. Example: Add a README with setup instructions

agent_id

string

Optional

Optional client-supplied agent ID in bc-<uuid> form, used for idempotency: re-sending the same agent_id returns 409 agent_id_conflict instead of creating a duplicate agent. Omit it to let Cursor generate the ID. Example: bc-00000000-0000-0000-0000-000000000001

auto_create_pr

boolean

Optional

Whether Cursor opens a pull request when the run completes. Defaults to false. Example: true

environment_name

string

Optional

Name of the saved Cursor-hosted environment, Team Pool, or self-hosted machine to run in. Only used when environment_type is set. A named cloud environment brings its own repositories, so do not also set repository_url. For type pool it defaults to the pool named default. Example: Release workspace

environment_type

string

Optional

Where the agent runs. 'cloud' uses Cursor-hosted VMs; 'pool' and 'machine' route to self-hosted workers. Omit it to use the default Cursor-hosted environment. Example: cloud. One of: `cloud`, `pool`, `machine`.

image_urls

array

Optional

Optional list of up to 5 HTTP or HTTPS image URLs that Cursor fetches and attaches to the initial prompt (PNG, JPEG, GIF, or WebP, up to 15 MB each). Example: ["https://example.com/screenshot.png"]

mode

string

Optional

Conversation mode for this run. 'plan' explores the code and drafts a plan before coding; 'agent' implements changes directly. Defaults to agent. Example: plan. One of: `agent`, `plan`.

model_id

string

Optional

Explicit model ID to run the agent with, as returned by cursor_models_list. Omit it to use the account's configured default model. Example: composer-2

model_params

array

Optional

Optional per-model parameters such as reasoning effort or max mode, only used when model_id is set. Each entry pairs a parameter id with a string value; use only parameters that cursor_models_list reports for the chosen model. Example: [{"id": "thinking", "value": "high"}]

name

string

Optional

Display name for the agent, 1 to 100 characters. Cursor derives one from the prompt when omitted. Example: Add README with setup instructions

pr_url

string

Optional

GitHub pull request URL to work on. The agent uses that PR's branches and starting_ref is ignored. repository_url must still be set to the PR's repository. Example: https://github.com/your-org/your-repo/pull/123

repository_url

string

Optional

GitHub repository URL the agent works in. Required when pr_url or starting_ref is set. Omit it, together with environment_name, to start an agent with no repository. Do not combine with a named cloud environment (environment_name with environment_type cloud). Example: https://github.com/your-org/your-repo

skip_reviewer_request

boolean

Optional

Whether to skip requesting the API key's user as a reviewer on the pull request Cursor opens. Only applies when auto_create_pr is true. Defaults to false. Example: false

starting_ref

string

Optional

Branch name or commit SHA in repository_url to start from. Ignored when pr_url is set. Example: main

work_on_current_branch

boolean

Optional

When false (the default), Cursor pushes commits to a new auto-generated cursor/... branch. When true, Cursor pushes directly to starting_ref, or to the PR's head branch when pr_url is set. Example: false

cursor_agent_get

Get agent

cursor_agent_usage_get

Get agent usage

cursor_agents_list

List agents

cursor_artifact_download_url_get

Get artifact download url

cursor_artifacts_list

List artifacts

cursor_environment_active_build_get

Get environment active build

cursor_environment_build_get

Get environment build

cursor_environment_builds_list

List environment builds

cursor_environment_create

Create environment

cursor_environment_get

Get environment

cursor_environment_history_list

List environment history

cursor_environment_secrets_list

List environment secrets

cursor_environment_update

Update environment

cursor_environments_list

List environments

cursor_me_get

Get me

cursor_models_list

List models

cursor_repositories_list

List repositories

cursor_run_cancel

Cancel run

cursor_run_create

Create run

cursor_run_get

Get run

cursor_runs_list

List runs

cursor_secrets_list

List secrets

cursor_team_secrets_list

List team secrets

Build your Agent

Same auth pattern across LangChain, OpenAI, Anthropic, and Google ADK.

import { ScalekitClient } from "@scalekit-sdk/node";
import { createReactAgent } from "@langchain/langgraph/prebuilt";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);

// Cursor tools scoped to this user
const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["cursor"], toolNames: [
    "cursor_agent_create",
    "cursor_run_get",
    "cursor_artifacts_list"] },
  pageSize: 100,
});

const agent = createReactAgent({ llm, tools });
await agent.invoke({ messages: [{ role: "user", content: "Start a Cloud Agent on acme/web to fix the failing login test" }] });

import OpenAI from "openai";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);
const openai = new OpenAI();

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["cursor"] }, pageSize: 100,
});

const res = await openai.chat.completions.create({
  model: "gpt-5",
  messages: [{ role: "user", content: "Start a Cloud Agent on acme/web to fix the failing login test" }],
  tools,
});

// Execute the tool call with the user's vaulted Cursor access token
await sk.tools.executeTool(res.choices[0].message.tool_calls[0], "user_123");

import Anthropic from "@anthropic-ai/sdk";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);
const anthropic = new Anthropic();

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["cursor"] }, pageSize: 100,
});

const msg = await anthropic.messages.create({
  model: "claude-sonnet-5",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Start a Cloud Agent on acme/web to fix the failing login test" }],
  tools,
});

// Tool call runs with the user's vaulted Cursor access token
await sk.tools.executeTool(msg.content, "user_123");

import { Agent } from "@google/adk/agents";
import { ScalekitClient } from "@scalekit-sdk/node";

const sk = new ScalekitClient(env.SCALEKIT_ENV_URL, env.SCALEKIT_CLIENT_ID, env.SCALEKIT_CLIENT_SECRET);

const { tools } = await sk.tools.listScopedTools("user_123", {
  filter: { connectionNames: ["cursor"] }, pageSize: 100,
});

const agent = new Agent({
  name: "coding_agent",
  model: "gemini-2.5-pro",
  instruction: "Cursor tools scoped to this user",
  tools,
});

await agent.run("Start a Cloud Agent on acme/web to fix the failing login test");

Try these prompts

Copy any prompt into your agent. Each maps directly to a Cursor tool. Click to copy, paste into your agent, done.

Cloud Agents

Copy the prompt

Copied

Start a Cloud Agent on acme/web to fix the failing login test.

Copy the prompt

Copied

List my Cloud Agents working on pull request #482.

Copy the prompt

Copied

Tell the docs agent to also update the README.

Runs and usage

Copy the prompt

Copied

Show the status of the latest run on the flaky-tests agent.

Copy the prompt

Copied

How many tokens did the refactor agent use across its runs?

Copy the prompt

Copied

Cancel the active run on the migration agent.

Environments and artifacts

Copy the prompt

Copied

List the saved cloud agent environments for our team.

Copy the prompt

Copied

Show the build history of the backend environment.

Copy the prompt

Copied

List the screenshots and logs the QA agent saved.

SEE HOW AUTH WORKS

Each user connects their own Cursor access token once, and Scalekit sends it with every call. Tokens stay vaulted, every call is scope checked, and every action is logged.

Authorize

Your user connects

Cursor

once. We tie it to their identity and the meetings they approved — no shared bot account, no org-wide access

Who:

user ‘A’

when:

Once per user

access:

Limited to user

Store

Their

Cursor

token lives in a vault scoped to them. User A's meetings are never reachable by an agent acting for user B, even on the same connection

vault:

encrypted

scope:

per-user

tokens:

auto-refreshed

Resolve

When your agent calls a

Cursor

tool, we fetch the right token server-side. It never touches your agent, never appears in the LLM context, never shows up in your logs

speed:

~40ms

check:

before every call

seen by:

nobody

Audit

Every

Cursor

tool call is logged — who triggered it, which meeting was fetched, what came back. 90 days of history, tied to the user who authorized it

history:

90 days

export:

SIEM-ready

logged:

every call

Test other agents

See the same per-user auth pattern across other developer tools and AI connectors.

Engineering Teams

DevOps assistant agent

Polls GitHub for failing checks and stale PRs, opens Linear issues for the ones that need work, and posts a daily digest to Slack. It acts as the engineer, not a shared service account.

Engineering Teams

Engineering standup agent

Pulls commits from GitHub and GitLab, tracks issue movement in Jira, and posts a per-engineer standup brief to Slack. Each engineer's activity is read on their own delegated OAuth.

Why Scalekit

Secure your agent's access. Connectors ship in minutes

01.

Shared tokens break per-user analytics

A shared Cursor access token looks fine in a demo. In production every agent run looks like one service account, and you cannot tell which user triggered it. Scalekit resolves the credential of the actual user who triggered the agent, never a shared bot.

// shared token
audit → bot_service_account

// scalekit
audit → user_abc ✓

02.

Authentication is not authorization

03.

Multi-tenancy is architectural

04.

Cursor today. Ten connectors tomorrow.

“Our agents act across Salesforce, Gong, Google Drive, and more, on behalf of every customer. Scalekit behind the scenes meant we can keep adding tools without ever rebuilding how credentials or tool calling work.”

Venu Madhav Kattagoni

Head of Engineering / Von

FAQs

Frequently Asked Questions

Does the agent access Cursor as the user or through a shared access token?

As the user. Each user connects their own Cursor access token once, and Scalekit sends it with every call. Cloud Agents, runs, and environment changes the agent makes are attributed to that user in your audit trail, not a shared service account.

Where is the Cursor access token stored?

In Scalekit's managed AES-256 token vault, namespaced per tenant. To rotate it, the user reconnects with a new access token. Revocation is a single dashboard action. Credentials never appear in prompts, logs, or LLM context.

Can I limit what the agent does in Cursor?

Yes. Filter by tool name in listScopedTools to expose only what you want, for example the list and get tools without cursor_agent_create and cursor_run_create, which start new agent work. Scalekit also enforces scope checks before every API call.

What happens when a user revokes Cursor access?

The connection is invalidated on the next tool call. Subsequent requests for that user fail closed with a clear error. Other users in the tenant remain unaffected. The event is logged for audit.

Can the agent read Cursor secret values?

No. cursor_secrets_list, cursor_team_secrets_list, and cursor_environment_secrets_list return secret names and versions, never values. The agent can see which secrets exist for the team, the user, and saved environments, while the values stay in Cursor.