Agenten-Kompass

AGENT PROFILES

Cursor

Cursor · Coding agents

Coding agent for understanding code, planning changes, developing, and testing.

Develop software

Cursor · AI development environment

Product and access

Cursor is an AI-powered code editor and development environment. Agent modes and Cursor CLI are interfaces for the same product.

01

Core features

8 features from provider sources · sources reviewed on 30 September 2026. Availability and limitations are listed directly below.

  • Agent, Ask and Manual modes

    Modes range from autonomous multi-part work to focused editing or read-only exploration.

    Source 2

  • Project rules

    Rules in .cursor/rules provide reusable, project-specific guidance.

    Source 3

  • MCP connections

    MCP connects external data sources and tools; agents can use configured tools.

    Source 4

  • Code and web search

    Agent tools support searching the codebase and web sources.

    Source 5

  • Terminal and file editing

    The agent can edit files and run terminal commands; approval and execution follow the selected mode.

    Source 5

  • MCP approvals

    MCP calls require approval by default; Auto-run changes this behavior.

    Source 4

  • Cursor CLI

    The CLI supports agent tasks, session continuation, MCP, and the same project rules.

    Source 6

  • Hooks during agent runs

    Scripts can add context at defined points, review or block actions, and log results.

    Source 7
02

Requirements and limits

‘Still open’ means the reviewed sources do not answer this point clearly.

  • Requirement

    Access and usage budget

    Cursor offers a free entry point and paid plans; agent usage is billed by model and included usage budget.

    Source 8

  • Requirement

    Approve MCP tools

    MCP calls require approval by default; automatic execution must be explicitly enabled.

    Source 4

  • Limitation

    Cloud hooks with narrower scope

    Cloud agents do not support all local hooks: IDE session, MCP execution, and tab completion hooks, among others, do not run there.

    Source 7

  • Limitation

    User hooks in the cloud

    Cloud agents do not load hooks from the local user directory; project hooks must be configured in the repository.

    Source 7
03

Practical tasks

Editorial examples based on documented features. No product tests performed.

01

Narrow down the cause of a bug

Your input
Repository, reproduction steps, and project rules
Possible result
The agent investigates affected areas and proposes a limited correction.
What to check
Review the diff and targeted checks yourself.
02

Prepare a code change for an issue

Your input
Issue, codebase, and approved MCP project tools
Possible result
Agent plans the work and creates code changes in the project.
What to check
Check tool approvals, changes, and local behavior.
EDITORIAL EXAMPLE

Investigate a reproducible error in a repository and prepare a small change with tests.

Your possible prompt

Read the project rules. Investigate [error], propose a small change, and run the relevant checks. Show me the diff and any remaining uncertainties.

Getting-started idea based on documented capabilities. Not an executed product test. Adapt the task, data, and permissions.

04

What you should know first

Cursor documentation reviewed. Agent modes, plan features, hooks, and MCP are documented; available models, quotas, and some hook types differ between the local IDE and cloud agent.

Requirements & limitations

Review code, tests, and changes. Specific access depends on the environment, plan, and project configuration.

Setup

Open the project and provide its rules.

Pricing model

Plan / model usage. Check current terms and included quotas with the provider.

The classification indicates the scope of effort. It is not a price or performance commitment.

05

Sources from the provider

Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.

Read profile as Markdown

How we assess the sources