Create Agents¶

Task agents are named background workers.

Each agent can have a different role, prompt, and model setup.

Ways To Create An Agent¶

Dashboard¶

Open the dashboard and go to the Agents tab.

The agent form supports:

name
role
personality
full system prompt
provider
model
base URL
API key
max concurrency

That means users do not need to edit config files just to give a task agent its own endpoint or API key.

CLI¶

arc agent create researcher \
  --role "Deep web research" \
  --personality "thorough, cites sources" \
  --model openrouter/anthropic/claude-sonnet-4-20250514

The CLI is convenient for quick creation, but the dashboard gives fuller visibility for agent setup.

Agent File¶

Agents are stored in:

~/.arc/agents/<name>.toml

Example:

name = "researcher"
role = "AI industry analyst"
personality = "thorough, evidence-driven"
max_concurrent = 1

[llm]
provider = "codex"
model = "codex-mini-latest"
base_url = "https://example.internal/v1"
api_key = "sk-secret"

system_prompt = """
You are a research agent.
Always cite sources and separate findings from assumptions.
"""

How Provider Inheritance Works¶

If an agent has its own [llm] block, it uses that.
If it does not, it falls back to the global default provider from ~/.arc/config.toml.

When To Give An Agent Its Own Full Prompt¶

Use system_prompt when you want tight control.

Use role and personality when a lighter setup is enough.

If system_prompt is present, it overrides the auto-generated prompt for that agent.