Cline

Cline asks before every file write and command, which sounds tedious until it saves you from a bad one, and on a real repo that happens weekly. Models are yours: your keys, ClinePass at $9.99 for open-weight models, or Ollama on your machine, so the bill and the privacy are decisions you make rather than accept. The daily trait is that you always know what it did, because you approved it.

Pick it if you want an agent you can read, approve, and point at any model. Pick Cursor for speed and Tab, where the editor does more and asks less.

The failure mode is fatigue. Cline asks before every action, the right default, which users eventually replace with auto-approve; the MCP settings schema has an autoApprove list built in, so the escape hatch ships with the product. Once you take it, there is no Docker sandbox, and a wrong command runs with your credentials in your checkout, the same as any other terminal agent minus the illusion.

Auto-approve reads and searches, never writes and shells. What it does right is transparency: Apache-2.0 source and readable prompts, so when it misbehaves you can find out why.

Cline's architecture is legible because the source is Apache-2.0. 1. Edits are applied per file with an approval step, so throughput is a function of the user's attention rather than the model's speed. 2. Plan and Act modes separate reasoning from execution, a decomposition other tools arrive at by accident and Cline documents on purpose, which means the plan can be reviewed as a document before any file changes.

No benchmark is listed, so claims are claims, and the design would be straightforward to measure. The observation: the approval step is both the safety mechanism and the bottleneck, and the docs do not pretend otherwise.

Cline runs the classic open-core play. The Apache-2.0 extension is free with roughly 67,000 stars, which is distribution nobody paid for. Monetization sits on top: ClinePass at $9.99 a month and a custom-priced Enterprise tier, thin layers over a free product. The moat is community, which cuts both ways: Roo Code was a fork, and the license invites the next one whenever the community disagrees with a roadmap decision.

Likely acquirer: an IDE vendor buying the user base, or a model provider buying a default client. Position: long the adoption, watch the gross margin, and read the license before assuming loyalty.

The demo is an extension that asks permission before each command, which reads well in a security review. The free path is our own keys: sixty engineers on our provider contract, no SSO, no audit log, and sixty local settings files nobody manages. The Enterprise path lists SSO/OIDC, SCIM, audit logs, RBAC and VPC deployment, which is the full checklist, but the price is custom, so the memo waits on a quote.

Onboarding is an extension install and a key. Approved with conditions: an Enterprise quote in writing and auto-approve disabled by policy across the fleet.

Cline is what I want an editor agent to be. The source is Apache-2.0 and the prompts are in the repo, so when it does something odd I read why and change it. The model picker takes Ollama, LM Studio, or any OpenAI-compatible endpoint, so my local box is a first-class provider. MCP servers live in the MCP settings JSON, and I version that file with the project.

A fork is a clone, and the community has already proven it. It is slower than the closed tools with a frontier model behind them; it is also mine. That is the point of the license.