Omni Config

Use this for runtime configuration and access management (not infrastructure bootstrap).

Auth and status

omni auth login --api-key <key> --api-url http://localhost:8882 --json
omni auth status --json
omni auth logout --json
omni status --json

Auth recovery

# Recover API key when keyValid shows "no" (requires local PM2 access)
omni auth recover --api-url http://localhost:8882 --json
# Generate a new key instead of recovering the existing one
omni auth recover --rotate --json

CLI config

omni config list --json
omni config get defaultInstance --json
omni config set defaultInstance <id> --json
omni config set format json --json
omni config unset defaultInstance --json

Provider schemas (quick reference)

Available schemas: agno, webhook, openclaw, claude-code, genie, a2a, ag-ui

For full provider management (list, get, create, update, delete, setup, test), see the omni-providers skill.

genie

Connects Omni to a Claude Code team inbox (Automagik Genie orchestrator).

omni providers create \
  --name "genie-prod" \
  --schema genie \
  --agent-name "omni-agent" \
  --target-agent "team-lead" \
  --team-name "omni-{chat_id}" \
  --json

Flags: --agent-name (required), --target-agent (required), --team-name (template with {chat_id}, {thread_id}, {sender_id}; default: omni-{chat_id})

claude-code

Runs Claude Code SDK locally as an agent backend.

omni providers create \
  --name "claude-dev" \
  --schema claude-code \
  --project-path /home/user/myproject \
  --max-turns 10 \
  --permission-mode acceptEdits \
  --model claude-opus-4-5 \
  --system-prompt "You are a helpful assistant." \
  --json

Flags: --project-path (required), --max-turns, --permission-mode (default|acceptEdits|bypassPermissions|plan), --model, --system-prompt

a2a

Agent-to-Agent protocol provider (Google A2A standard).

omni providers create \
  --name "a2a-agent" \
  --schema a2a \
  --base-url https://a2a-server.example/api \
  --api-key <key> \
  --json

ag-ui

Agent-UI protocol provider (CopilotKit AG-UI standard).

omni providers create \
  --name "agui-agent" \
  --schema ag-ui \
  --base-url https://agui-server.example/api \
  --api-key <key> \
  --json

openclaw / agno / webhook

# openclaw (WebSocket gateway)
omni providers create --name "openclaw-prod" --schema openclaw \
  --base-url wss://gateway.example/ws --api-key <key> --default-agent-id <agentId> --json

# agno / webhook (HTTP API)
omni providers create --name "agno-prod" --schema agno --base-url https://api.agno.com --api-key <key> --json

API keys

omni keys create --name "agent-key" --scopes messages:read,instances:write --instances <id1,id2> --json
omni keys list --status active --limit 50 --json
omni keys get <id> --json
omni keys update <id> --rate-limit 120 --expires 2026-12-31T23:59:59Z --json
omni keys revoke <id> --reason "rotation" --json
omni keys delete <id> --json

Server settings

omni settings list --json
omni settings list --category ai --json
omni settings get <key> --json
omni settings set <key> <value> --reason "ops update" --json

Service management

omni start --json
omni stop --json
omni restart --json
omni update -y --no-restart --json

Logs

omni logs
omni logs error --limit 50
omni logs --modules api,nats --follow
omni logs --process nats

Flags: [level] (debug, info, warn, error), --modules <modules>, --limit <n> (default: 100), --process [service] (default: api), --follow

Dead letters

omni dead-letters list --status pending --limit 50 --json
omni dead-letters get <id> --json
omni dead-letters stats --json
omni dead-letters retry <id> --json
omni dead-letters resolve <id> --note "fixed manually" --json
omni dead-letters abandon <id> --json

Payloads

omni payloads list <eventId> --json
omni payloads get <eventId> webhook_raw --json
omni payloads get <eventId> agent_response --json
omni payloads delete <eventId> --reason "cleanup" --json
omni payloads config --json
omni payloads config message.received --retention 30 --store-webhook true --json
omni payloads stats --json

Payload stages: webhook_raw, agent_request, agent_response, channel_send, error

Shell completions

omni completions bash
omni completions zsh
omni completions fish

Notes

• Prefer --json + jq for automation-safe parsing.
• omni providers setup is interactive; avoid it in automated agent flows.
• omni auth recover requires local PM2 access (run on the server hosting Omni).