How to use the Axiom Cost Control agent skill

What is an agent skill, and what does this one do?

An agent skill is a folder of instructions, reference files and scripts that an AI agent loads when a task matches the skill's description. This one is designed for Axiom users who want to lower ingest spend.

When it runs, the agent analyses Axiom query patterns to find columns and field values that are ingested but never queried, and ranks datasets by how much query value they return for their cost. It can then deploy a cost dashboard and create monitors that warn about daily ingest or query-cost spikes, and produce a report recommending which datasets to drop, sample or keep.

• Finds unused columns and field values in each dataset
• Ranks datasets by Work/GB (query cost per GB ingested)
• Deploys a cost control dashboard
• Creates three cost monitors, including an ingest guard against your contract limit
• Writes a report using the included analysis template

What you need before installing

The skill depends on Axiom and several other tools. Check these first, because every step relies on them.

The source lists these requirements: an Axiom account and deployment with an audit dataset (axiom-audit or similar), access to the axiom-history dataset, the companion skills axiom-sre and building-dashboards, the axiom-api and dashboard-* CLI tools, jq for JSON parsing, and a bash shell. The skill also uses network access and runs scripts.

• An Axiom account and deployment with an audit dataset
• Access to the axiom-history dataset
• Companion skills: axiom-sre and building-dashboards
• axiom-api and dashboard-* CLI tools
• jq for JSON parsing
• bash shell

Install the controlling-costs skill

Download the skill ZIP from our page and unzip it into one of two places. Use ~/.claude/skills/ to make it available across your projects, or a project's .claude/skills/ folder to limit it to that project.

In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

The README also mentions an install command, npx skills add axiomhq/skills, which the README says includes the skill dependencies. Install the companion skills as well, since this skill loads them in its first step.

• Option A: unzip into ~/.claude/skills/
• Option B: unzip into your project's .claude/skills/
• Option C: upload the ZIP in the Claude apps under Settings → Capabilities → Skills

Configure your Axiom deployment

The README says to create ~/.axiom.toml with your Axiom deployment details. The skill refers to deployments by name, such as <deployment>, so the name you configure is the one you will use in requests.

Before any monitors are created, the agent asks you for three things: the deployment name, the audit dataset name, and your contract limit in TB/day. The contract limit is required for the monitors.

• Deployment name
• Audit dataset name (try axiom-audit first; common alternatives are axiom-audit-logs-view and audit-logs)
• Contract limit in TB/day

How to trigger the skill with natural requests

You do not need to type script names. Ask for the outcome you want in plain language, and the agent picks the phases to run. The skill's description matches requests such as reducing Axiom costs, finding unused columns or field values, identifying data waste, or tracking ingest spend.

The skill maps each request to a set of phases. Phase 0 (checking for an existing setup) always runs first.

• "Reduce our Axiom costs" or "find waste" runs discovery, then optimisation
• "Set up cost control" runs discovery, dashboard and monitors
• "Deploy a cost dashboard" runs the dashboard only
• "Create cost monitors" runs the monitors only
• "Check for drift" checks the existing setup only

Examples of what to ask and what happens

Here is a typical workflow. You ask for waste to be found, the agent confirms your deployment and audit dataset, runs the baseline statistics, then analyses the largest datasets one at a time. For each dataset it checks unqueried columns, then field values such as app or service, and records an action of DROP, SAMPLE or KEEP.

For monitors, the agent lists your notification channels and asks which one to use. If you do not want notifications, it creates the monitors without one.

• Ask: "Find waste in our Axiom logs for deployment prod." The agent runs discovery and optimisation.
• Ask: "Set up cost control for prod with a contract of 5 TB/day." The agent creates the dashboard and monitors after confirming the limit.
• Ask: "Check whether our cost dashboard has drifted from the template." The agent runs only the setup check.
• Ask: "Remove the cost control monitors." The agent runs the cleanup steps.

Safety: scripts, network access and deleting data

This skill runs shell scripts and calls Axiom over the network. It creates and deletes dashboards and monitors in your live account. Read the scripts in the scripts/ folder before running them, and run them against a deployment you understand.

Do not drop data based only on the recommendations. Dropping a dataset can remove logs you need for incidents or compliance. Running create-monitors twice creates duplicate monitors, so delete existing ones first when re-deploying.

• Review scripts before the agent runs them
• Check which deployment the agent is targeting
• Review every DROP recommendation against incident and compliance needs
• Delete old monitors before re-running create-monitors

Tips for better results

Run any script with -h to see its full usage. Do not pipe script output to head or tail, because the source says this causes SIGPIPE errors. For ad-hoc APL queries, use the axiom-sre skill's axiom-query rather than calling the CLI directly.

• Run scripts with -h for full usage
• Do not pipe script output to head or tail (causes SIGPIPE errors)
• Install jq before starting
• Use axiom-sre's axiom-query for ad-hoc APL
• Note that dashboard filters use GB/month while scripts use TB/day