Glossary · Headless Coding Agents (CI Mode)

Headless Coding Agents (CI Mode).

Agents that run where no terminal watches.

Direct answer

Headless mode runs the agentic loop without a TUI — scripted prompts, piped I/O, machine-readable exit codes — for CI pipelines, cron jobs, and batch refactors. The discipline: deterministic scoping (plan files in, diffs out), approval policy per environment (--yolo only where revert is cheap), and memory-aware runs where past failures inform this run. Read-only plan mode previews changes without touching the tree.

In Anvaya

How We Implement It.

01Headless Coding Agents (CI Mode)anv --no-tui --yolo with exit codes 0/1/2 (denied), pipeable I/O, --plan-only read-only exploration, timestamped backups before mutations.

Questions

Asked About Headless Coding Agents (CI Mode).

Q

When is --yolo safe?

Where revert is one command (git-clean trees), scopes are narrow, and tests gate merges. Never on unreviewed production paths.

Q

How do headless runs stay cheap?

Spec-in/task-out discipline: fresh session per job with a written spec, packed memory instead of transcripts, quiet command flags.

Q

What should CI agents never do?

Force-push, broad auto-merge, secret-adjacent work, and unscoped network writes. Gate writes, log everything.

Stop Starting From Zero.

One binary. 11+9 Rust crates. 545 tests. Hand-written HNSW index. Three transport modes. Four providers, Ollama, Anthropic, OpenAI, Siemens. Zero API keys required to start. Mind remembers everything after the first session.