ZhikunCode

The deciding trait is that it lives somewhere rather than on something. You deploy it once and then reach it from a browser, including the one in your pocket, so checking on a long-running task from a train is ordinary rather than a stunt. Nothing is installed on the laptop you happen to be carrying.

You are the wrong buyer if you work on one machine and like it that way, because you would be operating a service for no benefit. Pick it when access matters. Pick a desktop tool when it does not.

The isolation is at the wrong granularity. The system is deployed as containers, and nothing documented gives an individual task its own boundary, which means several agents running at once share whatever the deployment can reach, including the shell and the repository. Scoped grants govern what an agent is allowed to ask for. They do not govern what a process can touch.

What it does right is version the workspace, so a confused run leaves a trail that can be unwound.

1. This is how a number should be published. The harness is the official one, the resolved count is given as 168 of 300, the patch generation rate is separated from the resolution rate, and the configuration names the model, the month, the closed six-tool set and the absence of network access and sub-agents.

2. Separating patch generation from resolution is the detail that matters, because it distinguishes failing to produce a patch from producing one that does not work. 3. The configuration is restricted, so the figure is a lower bound on the shipped product rather than a ceiling.

480 stars for something with this much engineering in it. There is no company named, no price and no hosted option, which means every user pays in operations rather than in money, and the author captures none of it. That is generous and it is not a structure that funds year three.

Moat: the published measurement, briefly, since credibility is transferable and slow to rebuild. Likely acquirer: none visible. Likely path: continued solo maintenance, or a hosted tier that changes the project's nature. Position: use it, and read the commit history before you depend on it.

An evidence chain per change is the most useful thing on this row for me. Screenshots, commands, console output, test results, video, network captures and diffs are collected so a reviewer can accept or reject on the record rather than on a summary, and that record is the thing an auditor asks for.

What is missing is identity: no SSO, no SCIM, and no mapping from an action to a person in my directory. Self-hosting answers residency and costs infrastructure. Approved with conditions: identity in front of the browser interface before anyone outside one team gets a login.

A core with no external dependencies is a claim I like more each year. It means the supply chain is the runtime and nothing else, no transitive package I have to audit, and a fork stays buildable long after the ecosystem around it moves on. MIT on top of that makes the fork mine.

Ollama is in the provider list, so the weights can live on my own machine, and MCP servers attach for tools. The one compromise is that it wants a deployment rather than a binary, and I run it as containers like everything else.