LOCAL RUNTIME / CONTROL PLANE

d1v CLI

Connect local development to your Agent workspace.

Install, authenticate, pair a device, and bind projects from the terminal. Execution stays visible in d1v.ai while local code stays on your machine.

Read full docs GitHub

SHA-256 verifiedmacOS & Linuxarm64 / x86_64

INSTALL / D1V-CLIREADY

01$ curl -fsSL https://d1v.gocan.ai/install/d1v-cli.sh | bash -s -- --install-skill auto

→

Resolving latest GitHub release

→

Verifying checksums.txt / SHA-256

~/.local/bin/d1v

AGENT SETUP

ALTbrew install d1vai/tap/d1vcargo binstall d1v-cli

02 / MINIMUM WORKING PATH

Verify the machine before starting an Agent.

These commands follow the actual runtime path. Each step has an observable result, so running a command is never mistaken for reaching the intended state.

Install and verify the binary

The installer detects OS and architecture and verifies the release before writing the binary. If Codex or Claude Code is already on PATH, it also installs the d1v skill for that agent.

$ curl -fsSL https://d1v.gocan.ai/install/d1v-cli.sh | bash -s -- --install-skill auto$ d1v --version$ d1v runtime doctor

The binary is available and runtime diagnostics report a concrete state.

Authenticate and inspect the account

After browser login, verify auth, project access, and the GitHub connection independently.

$ d1v auth login$ d1v auth status$ d1v project list --format json$ d1v github status

Authentication is valid and the expected projects are readable.

Pair and start the local runtime

Pair the device, start the Agent, then confirm its control-plane connection.

$ d1v agent pair$ d1v agent start$ d1v agent status

The device is online and Agent status is visible to the control plane.

Bind an existing project directory

Bind the repository path to an explicit project_id instead of inferring identity from a folder name.

$ d1v agent project bind --project-id <project_id> --path ~/work/my-app

New sessions can use this local runtime; existing sessions remain pinned.

03 / COMMAND SURFACE

The commands you need, without guesswork.

For automation, prefer `--format json`, then evaluate both exit status and structured fields.

AUTH01

Identity & configuration

d1v auth statusd1v config showd1v project list --format json

RUNTIME02

Device & Agent

d1v runtime doctord1v agent pair --code <code>d1v agent status

PROJECT03

Local projects

d1v agent project create --project-id <id> --name my-appd1v agent project import --path ~/work/my-app --project-id <id>

EXPOSE04

Temporary public access

d1v expose 3000d1v expose listd1v expose close <binding_id>

04 / SECURITY BOUNDARY

Local-first does not mean untraceable.

The local Agent opens an outbound-only connection. Credentials stay out of repositories, release archives are checksum-verified, and an offline local runtime fails explicitly instead of silently moving work to cloud.

Outbound-only connectionCredentials injected by a secret managerCommands and results remain traceable

PATHexport PATH="$HOME/.local/bin:$PATH"

DIAGNOSEd1v runtime doctor

SKILLd1v skill install --agent auto

UPGRADEd1v upgrade

REMOVEd1v uninstall

INSTALL & RECOVERY

05 / MACHINE-READABLE DOCS