Multi

The deciding trait is that there is no account. You install it, you connect a key, and you are working, with no email address handed over, no trial clock and no included-request allowance quietly counting down while you refactor. Anyone who has abandoned a tool at the registration screen will recognise how much friction that removes.

What you should notice is that free with no sign-up is a decision somebody can revisit. Pick it for the editor you already use. Pick an open project if you need to know the terms will hold.

Category-level consent is the wrong granularity for the dangerous category. Reads and todo updates are fine to wave through, and grouping command execution into the same mechanism means the setting that makes the tool pleasant is the setting that removes the last check before something irreversible. Background tasks make it worse: the prompt you would have read is not on screen.

What it does right is separate the categories at all. Most tools here offer one switch, and a user who wants reads automatic and writes reviewed can have exactly that.

1. Treating a session as a state that can be forked is the right abstraction for a process whose output is uncertain, because it makes an unsuccessful attempt cheap to abandon instead of expensive to unwind. That is a design decision, and it is stated as one.

2. The parallel-workstream claim is weaker than its wording. Several tasks running at once is concurrency, not coordination, and no component is described that reconciles what two of them did to the same file. The distinction should be in the marketing, and is not.

Give it away, capture the workflow, charge the employer later. The mechanics are sound and the structure is visible in the licence itself, where individual use is free and anything commercial is a conversation, which is a funnel disguised as generosity. It also means the terms your company operates under are not published anywhere.

Moat: habit, and habit is real once an editor extension is in daily use. Likely acquirer: an editor vendor wanting a multi-provider layer. Position: enjoy the free tier, and read the commercial terms before anyone bills a client for work done inside it.

The free tier is written for people, not for companies, which means my sixty engineers are on the wrong side of the line the day they open a work repository. An unpublished commercial rate is a rate I cannot forecast, and the licence question alone will occupy my legal team longer than the evaluation would.

Beyond that there is no console, no directory integration, no audit export and nothing that runs in a pipeline, so it improves individual output and produces no measurement I can report. Not yet: published commercial terms first, then a security questionnaire.

Grudging respect for the provider layer. Riding an existing subscription instead of demanding another key is the pragmatic thing nobody else bothers with, and profiles that bind a provider and a model together mean switching mid-task is one selection rather than a settings excursion. A local runtime sits in that list as a peer.

Then the walls close in. I cannot read it, cannot fork it, and there is no protocol client, so every server I run is invisible to it. Wide model freedom, zero tooling freedom, which is an odd pair of choices.