Configuration File

config.yaml holds what the flags set. Change it with a2a-serve config set, a2a-serve channel add or a2a-serve config edit. Secret values are references such as secret:TELEGRAM_BOT_TOKEN.

name: coder
agent: opencode
command: [cloudcode, acp]        # optional: replaces the profile's command
sandbox: docker                  # or local
workspace: /home/u/src/app       # optional: default the directory holding .a2a
mounts: [/home/u/docs:/docs:ro]  # besides the workspace
model: openrouter/~google/gemini-flash-latest  # optional: the agent's own model id
secrets: [OPENROUTER_API_KEY]    # passed to the agent as environment variables
port: 46757                      # localhost port, chosen on the first run
network:
  private: tailscale
  public: tailscale
policy:
  privateAccess: paired          # default everyone
  publicAccess: paired           # or anyone
  allowEmails: [mom@example.com, "*@example.com"]
channels:
  telegram: {token: secret:CODER_BOT_TOKEN}
approvals:
  allow: [read, search]
  ownerChat: telegram:anna
mcp:
  disabled: [kubernetes]         # servers of .mcp.json the agent doesn't get
limits: {memory: 8g, cpus: "4", pids: 1024}
taskTTL: 30d                     # tasks idle this long are deleted
auth:
  googleClientID: 1234.apps.googleusercontent.com
  googleClientSecret: secret:GOOGLE_CLIENT_SECRET
backup:
  to: gs://my-bucket/agents/coder
  key: secret:BACKUP_KEY
  interval: 1h
  retention: 7d

Keys

The table lists the key each flag sets. Keys are camelCase and nest at each dot. Files with the snake_case keys of older versions still load, and a2a-serve doctor --fix renames their keys.

Flag	Key	Commands
--agent	agent	config set
--name	name	start <agent>, config set
--description	description	start <agent>, config set
--command	command	start <agent>, config set
--model	model	start <agent>, config set
--image	image	start <agent>, config set
--sandbox	sandbox	start <agent>, config set
--workspace	workspace	start <agent>, config set
--mount	mounts	start <agent>, config set
--secret	secrets	start <agent>, config set
--env	env	start <agent>, config set
--port	port	start <agent>, config set
--private	network.private	start <agent>, config set
--public	network.public	start <agent>, config set
--private-access	policy.privateAccess	start <agent>, config set
--public-access	policy.publicAccess	start <agent>, config set
--allow-email	policy.allowEmails	start <agent>, config set
--tailscale-control-url	network.tailscaleControlURL	config set
--tailscale-authkey	the secret TS_AUTHKEY in state/secrets	config set
--relay-token	the secret RELAY_TOKEN in state/secrets	config set
--token	channels.telegram.token	channel add telegram
--token	channels.discord.token	channel add discord
--email-address	channels.email.address	channel add email
--email-imap	channels.email.imap	channel add email
--email-smtp	channels.email.smtp	channel add email
--email-username	channels.email.username	channel add email
--email-password	channels.email.password	channel add email
--email-mailbox	channels.email.mailbox	channel add email
--email-trusted-authserv	channels.email.trustedAuthServ	channel add email
--email-oauth-token-url	channels.email.oauthTokenURL	channel add email
--email-oauth-client-id	channels.email.oauthClientID	channel add email
--email-oauth-client-secret	channels.email.oauthClientSecret	channel add email
--email-oauth-refresh-token	channels.email.oauthRefreshToken	channel add email
--allow-tools	approvals.allow	start <agent>, config set
--deny-tools	approvals.deny	start <agent>, config set
--disable-mcp	mcp.disabled	start <agent>, config set
--owner-chat	approvals.ownerChat	start <agent>, config set
--memory	limits.memory	config set
--cpus	limits.cpus	config set
--pids	limits.pids	config set
--google-client-id	auth.googleClientID	config set
--google-client-secret	auth.googleClientSecret	config set
--backup	backup.to	config set
--backup-key	backup.key	config set
--backup-interval	backup.interval	config set
--backup-retention	backup.retention	config set
--task-ttl	taskTTL	config set