Configure .archal.json
Create.archal.json in your project root:
| Field | Required | Description |
|---|---|---|
agent | Yes | Shell command to run your agent. Can be a string or an object with command, args, and env. |
title | No | Display name for this project. |
twins | No | Which twins to start. If omitted, inferred from the scenario. |
scenarios | No | Array of scenario file paths relative to .archal.json. |
seeds | No | Per-twin seed names, e.g. { "github": "small-project" }. |
agentModel | No | LLM model for the agent (e.g. claude-sonnet-4-6). |
model | No | Evaluator model (e.g. gemini-2.5-flash). |
runs | No | Default number of runs per scenario. Default: 1. |
timeout | No | Default timeout per run in seconds. Default: 180. |
agent field also accepts an object for more control:
Run a task
The quickest way to test is an inline task:.archal.json with an agent field, Archal uses that command. Otherwise it runs the task directly.
Run a scenario
For repeatable tests, write a scenario file and pointarchal run at it:
How the proxy works
When you pass--proxy, Archal starts a local TLS proxy that intercepts your agent’s HTTPS traffic:
Supported runtimes
The proxy is language-agnostic (standard HTTPS CONNECT), but CA certificate injection varies:| Runtime | CA injection | Notes |
|---|---|---|
| Node.js | NODE_EXTRA_CA_CERTS | Automatic |
| Python | REQUESTS_CA_BUNDLE, SSL_CERT_FILE | Automatic |
Intercepted domains
| Service | Domains routed to twin |
|---|---|
| GitHub | api.github.com, uploads.github.com |
| Google Workspace | gmail.googleapis.com, calendar.googleapis.com, drive.googleapis.com, sheets.googleapis.com, people.googleapis.com, oauth2.googleapis.com |
| Slack | slack.com/api/* |
| Stripe | api.stripe.com |
| Jira | *.atlassian.net |
| Linear | api.linear.app |
| Supabase | *.supabase.co |
| Discord | discord.com, discordapp.com |
Without the proxy
If your agent already talks to services via MCP tools (not raw HTTP), you don’t need--proxy. Archal passes MCP endpoints and REST base URLs as environment variables. Your agent uses those directly.
View results
Results print to the terminal. They also appear at archal.ai/dashboard.Related
- Writing scenarios
- archal run reference
- Twin sessions for persistent twins outside the run lifecycle
- CI integration