ACP Client for VS Code

The deciding trait is that switching agents stops being a project. The extension hosts whichever one you want inside the editor you already use, so trying a different agent next week means picking it from a list instead of learning a new interface and a new set of keybindings. Given how fast this category changes, that is worth more than any single agent's advantage.

Only one runs at a time, and it adds nothing to what the agent can do. Pick it if you keep wanting to try the new thing. Pick a vendor's own extension if you have already decided and you like their panel.

Auto-approve is the setting to read twice. The extension serves the hosted agent's file reads and writes against your workspace and runs the commands it asks for, and the policy governing all of that is configurable, which means it can be configured wrongly once and then never thought about again. The failure is silent by design: approval is the only gate and you have turned it off.

The compensating design is that the extension does nothing on its own. It has no model and no loop; everything that goes wrong here was asked for by an agent you chose, which narrows the blame but does not stop the edit.

1. Every ACP message can be inspected in a traffic log. That is an unusual and valuable property: the protocol between the client and the agent is normally opaque, and here the JSON-RPC exchange is available to the person debugging it, which makes a disagreement about whose fault a failure is answerable rather than rhetorical. 2. It also makes the extension a usable instrument for studying agent behaviour, which is not what it was built for.

2. There is nothing to benchmark, since the extension contributes no capability of its own. Its correctness is protocol conformance and that is checkable.

The value of this extension is entirely borrowed. It is worth exactly as much as the protocol is adopted, and if the large vendors decide their own panel is the product, the list of hosts it can speak to stops growing and this becomes a compatibility shim for whoever is left.

Moat: the author's reputation and an install base already in place, neither of which is defensible. 372 stars for a free extension is respectable and unmonetisable. Likely path: it remains a well-maintained utility. Likely acquirer: none; extensions are not bought, their authors are hired. Position: use it, expect the list to change.

This is the cheapest rollout on the board. One marketplace identifier goes into the extension allowlist we already maintain, and sixty engineers have it the next time the editor restarts; no new application, no packaging, no endpoint exception. That is a memo, not a project.

What it does not solve is spend. Each engineer still brings their own agent and their own subscription, so I get no consolidated invoice and no view of what is being used. There is no audit trail and nothing runs in a pipeline. Approved with conditions: a standard configuration pushed with the extension rather than sixty personal ones.

MIT, and the important line is that the agent list is not a list. Anything speaking the protocol can be configured by hand, which means the thing I am writing this month is hostable in the same panel as the commercial ones, with no plugin API to learn and no review to pass.

The per-session selectors are advertised by the agent rather than hard-coded by the extension, so a host that grows a new mode does not need this to ship a release. There is no MCP client of its own, which is correct: my servers belong to the agent I launched, not to the window it draws in.