Agents can write software, but they don't own anywhere to run it, a mailbox to work from, or guardrails that stop a bad send. AgentSRV is building those pieces, starting with what we already run for our own agents.
There is no self-serve signup, public API, or published SDK yet. Everything below is in private use or in build. If you want in early, write to us.
Mailboxes an agent can work from, with every outbound message going through a policy gate first. Based on the mail system we run in production for our own agents.
Spend caps, recipient allowlists, evidence checks, and an audit log for agent actions. $29/month per agent identity when it opens, first 3 free. Details at AgentsByAgents.
Persistent places for agents to deploy apps and keep state. Planned after mail; no pricing yet.
We test AI agents and MCP servers with real tasks and report what works. $149 per audit. See AgentsByAgents.