Guides for running agents in production
Practical guides to silent agent failures, cost monitoring, human approvals, and output review. Explore the glossary and build a supervised rollout checklist.
12 guides
How to set a spending limit on an AI agent
How to put a hard spending limit on an AI agent: per-run cost, token and runtime caps, where to enforce them, and what must happen when one is hit.
Read the guide →How to stop a runaway AI agent
How to spot an AI agent stuck in a loop, stop it with limits or a kill switch, end its whole process tree, and design runs so a stop is safe.
Read the guide →How to control which MCP tools an agent can use
How to restrict the MCP tools an AI agent can call: least privilege, allowlists, call budgets, approval of exact calls, clean refusals and logging.
Read the guide →How to detect silent failures in scheduled AI agents
Why scheduled AI agents stop without an error, and how to catch it: expected-schedule checks, stale heartbeats, stalled runs, output checks and alert routing.
Read the guide →How to design human approval for AI agent actions
Which AI agent actions need human approval, how to approve the exact action rather than an intent, and what should happen when a request is denied or ignored.
Read the guide →How to monitor what your AI agents cost, per run and per agent
Where AI agent cost figures come from, how token types and price tables change them, and how to track LLM cost per run, per agent and per month.
Read the guide →How to monitor Claude Code and Codex usage across a team
Monitor Claude Code and Codex across a team with OpenTelemetry: what to collect, what to keep out, how to read cost, and how to limit unattended runs.
Read the guide →OpenTelemetry for AI agents: what to trace and how
How to trace AI agents with OpenTelemetry: which spans to record, GenAI semantic conventions, what to keep out of traces, and exporting over OTLP/HTTP.
Read the guide →A rollout checklist for putting your first AI agent into production
A step-by-step checklist for deploying an AI agent to production: define a good result, test failure paths, set limits and approvals, then decide go or no-go.
Read the guide →Agent monitoring and agent supervision answer different questions
Monitoring shows an AI agent is running. Supervision checks that its work arrived, was good and was allowed. How the two differ and how they fit together.
Read the guide →Approving an AI agent’s action does not prove it happened
A person approving an AI agent’s action does not show it happened correctly. How to record authorization, receipts and independent verification separately.
Read the guide →What an AI agent audit trail should record
What an AI agent audit trail should record: runs and steps, claimed versus measured data, who approved each action, policy and credential changes, retention.
Read the guide →Plan your first agent pilot
A practical checklist for reporting, failure testing, outcome review, and guarded actions. Use it in your browser or download it for your team.
Connect your first agent
Create a free workspace, connect the agent you already run, and watch its next run arrive. Free covers 3 agents, with no card required.