Sign in

Skip to content
Teev Docs
Console

Documentation

Which AI coding tools actually work on your code.

Teev builds a private benchmark from your own merged pull requests, then runs model and harness combinations against those tasks in your own cloud. This site will document how to set that up and how to read the results.

Documentation is being written. Two pages are published below; the rest are listed so you can see what is coming.

Where your data lives

Teev runs as two parts. A hosted control plane handles sign-in, accounts and scheduling. A runner you deploy into your own cloud account does the actual work: reading your repositories, building environments, running agents and verifying results.

Never leaves your account

  • Source code and repository snapshots
  • Your GitHub token
  • Your model provider API keys
  • Patches, diffs and agent transcripts
  • Container logs and full run artifacts

Reported to the control plane

  • Account and user identity, sign-in events
  • Repository names and task identifiers
  • Benchmark definitions — rules, targets, model selections
  • Result summaries — pass rates, timings, token counts, cost
  • Runner health and capability metadata

The claim is your source and your compute never leave your account — not that we see nothing. The runner reports outward over HTTPS; nothing reaches into your account, and you hold its source and control its egress rules.

What the install command does

The console gives you a one-time command for the machine that will host your runner. It downloads a shell script from teev.ai and runs it as your user. That script is the entire installer, and this is what it does, so you can read it against the copy in front of you. 

curl -fsSL https://teev.ai/install -o teev-install.sh
TEEV_BOOTSTRAP_TOKEN=<one-time-token> sh teev-install.sh

Downloading and running are two steps deliberately. Piping the script into sh takes over the terminal's input, and the installer asks questions, so a piped install hangs rather than prompting. The gap is also where you get to read the file.

What it changes on the machine

  • Creates ~/teev-runner and unpacks one release archive into it
  • Writes the runner's identity and state under ~/teev-runner/.teev
  • Starts the runner — Docker Compose on Linux, a launchd agent using Docker Sandboxes on macOS
  • Builds the runner image, which pulls base images and packages from the usual public registries
  • No sudo, nothing installed system-wide, no edits to your shell profile or PATH

What it sends out

  • One request to the control plane as it starts, carrying the bootstrap token, so the console can show that setup began
  • Downloads from teev.ai: the release archive, plus its checksum file only if you asked for a non-default version
  • Enrollment, where the runner registers and receives its long-lived credential
  • Nothing describing your machine, your repositories or your environment — from then on the runner reports only what the section above lists

Verifying the release

The published installer pins the runner version and the SHA-256 of that version's archive, as a literal near the top of the file. It refuses to unpack a download that does not match. To check the pin by hand before you run anything: 

grep '^TEEV_RELEASE_SHA256=' teev-install.sh
curl -fsSLO https://teev.ai/releases/runner-v<version>.tar.gz
shasum -a 256 runner-v<version>.tar.gz

Before it downloads anything the installer checks the host and stops with a plain message if something is missing: macOS or Linux (on Windows, WSL2), Docker running — or Docker Sandboxes on macOS, which it requires be set to deny network access by default — and Node 22 or newer. It warns, but continues, below 20 GB of free disk.

Running the same command again upgrades an existing runner in place. Your .teev state, model provider credentials and packaged work are left alone. The bootstrap token is single-use: on a machine that is already enrolled, a live token moves that runner to the account that issued it, and a spent one is ignored so the command is safe to re-run.

The installer is not code-signed, and a signature published next to the file it signs would not tell you much more than the file does. The trust you extend is to teev.ai over HTTPS at the moment you download the script — which is why the hash that verifies the release ships inside the script you just read rather than beside the archive, and why the script is written to be read.

Planned

Not yet written. Listed so the shape of the documentation is visible.

  1. Getting startedConnect a repository, screen candidate tasks, build your first benchmark.
  2. Deploying a runnerTemplates for AWS, Azure and plain Docker, and what each one provisions.
  3. Task selectionHow pull requests become tasks, and the rules that decide eligibility.
  4. Execution targetsHarness, provider, model, effort and trial count — how a comparison stays fair.
  5. Reading resultsPass rates, calibration, cost per task, and what a result does not tell you.
  6. Security and operationsThe full data-handling model, retention, and access control.
Teev · console.teev.ai ·