Control plane for AI agents
See, govern, and stop all your agents from one place. Dashboard, health monitoring, policy enforcement, and kill switch for production AI agents.
$ axme agents list
AGENT HEALTH COST/DAY INTENTS
● agent://acme/prod/invoice-bot healthy $2.40 1,247
● agent://acme/prod/support-ai healthy $8.10 3,891
● agent://acme/prod/data-analyst degraded $14.30 892
● agent://acme/prod/email-sender unreachable $0.00 0
$ axme agents kill agent://acme/prod/email-sender
Agent killed. All intents blocked. Resume with: axme agents resume
Four things you need for agents in production
63% of companies can't enforce purpose limitations on their agents. 60% can't terminate a misbehaving one. You shouldn't be one of them.
Dashboard
All your agents on one screen. Who is alive, who crashed, how many tasks each processed, how much each spent on LLM calls. Real-time.
Health Monitoring
Automatic liveness checks every 30 seconds. Agent goes down - you get an alert in Slack, email, or webhook. See degradation before users complain.
Policy Enforcement
Define what each agent can do: allowed actions, spending limits, data rules, operating hours. The platform enforces them - not the agent itself.
Kill Switch
One button to instantly isolate a misbehaving agent. No intents in, no intents out. Takes effect in under 1 second. Reversible when the issue is fixed.
Answer honestly
How many AI agents do you have running right now?
What's the error rate for each one?
How much did they spend on LLM calls this week?
Can you stop a specific agent in under 3 seconds?
If you can't answer all four - you need Agent Mesh.
Three steps to production-grade agents
Register your agents
Add the SDK to your agents. They auto-register in the mesh and start sending heartbeats.
from axme import AxmeClient
client = AxmeClient(api_key="axme_sa_...")
client.mesh.start_heartbeat()Set policies
Define what each agent is allowed to do. Cost limits, action types, operating hours, approval gates.
axme agents policy set agent://acme/prod/email-sender \
--max-cost-daily 10 \
--allowed-types "email.send" \
--require-approval "email.send_bulk"Watch and govern
Real-time dashboard shows health, costs, policy violations. Kill switch when something goes wrong.
axme agents list # see all agents + health
axme agents kill <agent> # instant isolation
axme agents resume <agent> # bring it backWorks with any framework
Agent Mesh is protocol-first, not framework-dependent. If your agent can make HTTP calls, it works with Mesh.
Free for up to 50 agents
Register your agents, see them in the dashboard, set policies, use the kill switch. No credit card. No time limit.