Ravi

1code

What Is 1code?

1code (by 21st.dev) is an open-source desktop app and orchestration platform for coding agents. It treats Claude Code and Codex as first-class workers — not copilots — and gives each agent its own isolated git worktree so parallel sessions never conflict. Key features include diff previews, a built-in git client, a Kanban board for tracking parallel work, and a REST API for programmatic task dispatch.

The standout capability: background agents that run in isolated cloud sandboxes even when your laptop is closed. These agents can be triggered by GitHub comments, CI failures, or Linear tickets. They clone repos, run tests, commit, push branches, and open PRs without human intervention.

Where Ravi Fits

Running agents autonomously at this level surfaces identity and credential problems quickly. Ravi solves four specific friction points that 1code users hit in practice.

1. Secrets vault instead of dotfiles

1code agents need API keys for Claude, OpenAI, GitHub, Linear, Slack, and more. Without a structured solution, teams manage these as environment variables or config files scattered across machines and sandboxes. Ravi’s encrypted secrets vault gives agents a consistent, secure place to retrieve credentials at runtime — no dotfile juggling, no plaintext tokens committed to worktrees.

2. Per-agent GitHub identity

When multiple parallel agents each open PRs against the same repo, you need distinct committer identities to keep authorship traceable. Ravi provisions unique email addresses per agent. Set each one as the agent’s git commit email and every PR has a clear, auditable author without credential sharing.

3. Cloud sandbox verification

Background agents frequently need to register for services or receive OTP codes during setup — SMS verifications, email confirmations, account activations. Ravi’s provisioned phone numbers and email addresses let agents complete these flows autonomously inside cloud sandboxes, without routing through your personal contact info.

4. OAuth token storage

Connecting 1code to GitHub, Linear, or Slack involves OAuth handshakes that produce access tokens. Ravi stores those tokens securely in its secrets vault and surfaces them to agents at runtime — so the token lives in one place, accessible to any agent that needs it, and rotatable without hunting through config files.

Integration Guide

Option 1: Environment variables at session startup

Configure 1code to inject Ravi-sourced secrets as environment variables when starting an agent session. In your 1code configuration:

{
  "env": {
    "GITHUB_TOKEN": "${ravi:GITHUB_TOKEN}",
    "OPENAI_API_KEY": "${ravi:OPENAI_API_KEY}",
    "LINEAR_API_KEY": "${ravi:LINEAR_API_KEY}"
  }
}

Agents read these from the environment as normal — no code changes required.

Option 2: Ravi as MCP plugin

The cleanest integration is a Ravi MCP server deployed in 1code’s plugin marketplace. Agents running in any 1code session can call Ravi tools directly from their tool loop:

  • ravi:get_secret("GITHUB_TOKEN") — retrieve a stored API key
  • ravi:get_email() — get the agent’s provisioned email address
  • ravi:get_phone() — get the agent’s provisioned phone number for OTP flows

This eliminates env var juggling entirely. Agents fetch the credential they need, when they need it, with full audit trail.

Setting up per-agent git identity

For each parallel agent in 1code, configure a distinct Ravi identity and use its provisioned email as the git committer:

# In the agent's session startup script
git config user.email "$(ravi get-email)"
git config user.name "$(ravi get-identity-name)"

Each agent’s commits are now attributed to its own Ravi identity — distinguishable in GitHub’s PR view and the commit log.

OTP handling in background agents

When a background agent needs to complete an email or SMS verification:

  1. The agent uses its Ravi-provisioned email or phone to initiate the verification
  2. It calls ravi:check_inbox() or ravi:check_sms() to poll for the code
  3. It extracts the OTP and completes the flow

No human is needed in the loop. The agent handles the full auth handshake autonomously.

Why This Matters

1code’s value proposition is autonomous agents that work while you sleep. Ravi makes that promise hold up in practice: agents that can authenticate, verify, and operate with real credentials — not workarounds. The combination is particularly powerful for background agent runs where there’s no human available to paste a code or approve a login.