coi

Give the agent a machine. Just not yours. Each AI coding agent gets its own isolated machine with root, Docker, and systemd - active defense detects and stops threats automatically.

Runssource: GitHubhomepageGoMITcommit a6781344bd1c

Go, MIT licensed. The project labels itself: agentic ai, ai tools, anthropic, claude, claude code, code sandbox, codex and coding assistant.

coi runs. An Argusic agent installed it in 10 minutes and hit 5 errors and fixed every one without help on a clean machine with no GPU, and the whole session was recorded.

Coi binary builds from source, runs 'coi version' (Coi vdev) and 'coi --help' (lists 27 commands), passes all 23 Go unit test suites with -race enabled, and integration-tier compilation vet passes.

What the agent ended up with on a clean machine, in its own words. How this is measured

Measured by Argusic on a fresh machine every time. Every number links to its evidence. Argusic Score 96 of 100 (the mean of this project's run scores).

At a glance

verdict
Runs
Argusic Score
96 / 100
cost of the verifying run
$0.07 (measured)
recorded runs
2
last tested
stars
791
forks
69
open issues
40
watchers
4
size
12 MB
created
last push

Subject data from GitHub, linked at the top of this page, refreshed . Test data by Argusic (CC BY 4.0); every number links to a run page with the full log, the recording, and their sha256 hashes.

What was measured

From the run that decided this verdict. Every line is a recorded value, not an estimate.

Time lost to each failure while testing coi, drawn from the recorded run
Every bar is a failure the agent recorded, with the time it cost. Drawn from this run, not from anywhere else.

What went wrong, in the agent's own words

The agent's own words, worst first, never rewritten into advice.

  • TestInstallSh_NftAutoSetup_NonInteractive failed because nft not in PATH (pkg_install tried sudo apt-get)3 minutes
  • Go 1.27.4 needed (go.mod requires 1.26), not in PATH2 minutes
  • cgo build failure: systemd/sd-journal.h not found (libsystemd-dev not installed)2 minutes
  • make build needs embedded files: internal/image/embedded/dummy and internal/config/embedded/default_config.toml1 minute
  • Python CLI tests fail because incus is not installed (13 failed due to incus: executable not found)
Install time
10 minutes
Cold machine to finish
10 minutes
Errors hit and fixed
5 hit, 5 fixed with no human help
How the result was proved
go test -short -race ./... passed all 23 Go packages; binary runs and prints version/help/core commands; go vet -tags integration ./... passes; nft --version works with extracted binary
Model tokens used
83,002
Exact commit tested
a6781344bd1c
Container image
sha256:cdd920bce783

What was not measured

The other half, so you can see exactly how far this evidence goes. A gap is printed as a gap and never filled in.

Peak GPU memory
this ran on a CPU machine, so there was no GPU to sample
Behaviour over a long run
each attempt uses the smallest documented settings and a fixed time budget, so this measures whether it runs, not how it holds up under load

The end of the recording

The last lines of the terminal session that decided this verdict, exactly as they were recorded.

ok    github.com/mensfeld/coi/internal/terminal  1.027s
ok    github.com/mensfeld/coi/internal/timing  1.017s
ok    github.com/mensfeld/coi/internal/tool  1.078s
ok    github.com/mensfeld/coi/internal/tool/credentials  1.020s
ok    github.com/mensfeld/coi/internal/vmhost  1.023s
ok    github.com/mensfeld/coi/schema  1.204s
All 23 Go packages pass with `-race` enabled. Let me clean up the workspace.
2026-11-22T13:26:15.825646Z ERROR codex_core::tools::router: error=exec_command failed: CreateProce…
Now let me compile the final report.
tokens used
83,002
Now let me compile the final report.

Replay the whole session, every command from a clean machine to this point.

How it was tested

One agent, one container, no human help. It installs the project the way its own documentation says to, runs it, and fixes what breaks. Everything below is recorded as it happened: the terminal session, the log and the exact commit. The full method.

Strengths and limits

Measured facts, not opinions. How this is written.

What went well

  • Reached a running state on a clean machine, with the session recorded.
  • Recovered from all 5 errors without a human stepping in, which says the failures are documented well enough to solve.
  • Ran without a GPU, so it does not need one to start.
  • Licensed MIT, as reported by its host.
  • Measured 2 times, so the result is not a one-off.

What did not, or is not known

  • Took 10 minutes to install, slower than the median of the 38 comparable projects Argusic has measured.

Of the 39 agentic-ai projects Argusic has installed and timed, coi was the 26th fastest to reach a running state, and 31 of 39 reached one at all.

What it is a reasonable choice for

  • Trying it on a laptop or a small server: it reached a running state without a GPU.
  • A quick evaluation: it was running 10 minutes after the clone.

Also tested, in the same area

Every one of these was installed and run by Argusic on a clean machine. Nothing appears here that was not tested.

Run history

tested byattemptoutcomescorecost USDwhen
Argusic Runner2/3Runs100.000.07
Argusic Runner1/3Runs with mocks92.000.05

Topics (from GitHub)

agentic-aiai-toolsanthropicclaudeclaude-codecode-sandboxcodexcoding-assistantcontainer-securitycontainersdeveloper-toolsincusllm-securityllm-toolslxcopencodeopencode-aisandboxsandboxing-toolsecurity

Embed the badge

Markdown for the project README. It links back here; terms on the terms page.

[![Tested by Argusic](https://argusic.com/badge/coi.svg)](https://argusic.com/subject/coi)

Questions

Does coi run?
Yes. coi runs. Argusic installed and launched it on a clean machine in 10 minutes, hitting 5 errors on the way, and recorded the session.
How did Argusic test coi?
On a fresh, disposable machine, with every command recorded and the repository pinned at commit a6781344bd1c. 2 attempts are recorded, and the full method is on the methodology page.
What did it cost to test coi?
The run that produced this verdict cost $0.07: real compute and model cost, not a list price per million tokens. It is recorded on that run's page.
How long does coi take to install?
10 minutes in the run that decided this verdict, from a clean machine to a running state. That is one measured attempt, not an average.
Does coi need a GPU?
Not to start. Argusic reached a running state on a machine with no GPU.
What goes wrong when installing coi?
5 things broke in the recorded run, and 5 were fixed without human help. Each one, and the time it cost, is listed on this page.
How does coi compare with the alternatives?
Of the 39 agentic-ai projects Argusic has installed and timed, coi was the 26th fastest to reach a running state, and 31 of 39 reached one at all.
Where is the evidence for coi?
All 2 recorded runs are on this page, each linking to its full log and terminal recording, stored with a sha256 fingerprint so it cannot be quietly altered.

Discussion