New Axis Runtime v2 is now generally available

Ship AI agents your engineering team can actually trust.

Axis is the control plane for production AI agents — orchestration, evals, guardrails and observability in one runtime, so you can ship agentic features without shipping the incidents that come with them.

npm install @axis/sdk — no credit card required
Overview
Agents
Traces
Evals
Guardrails
Settings
Support Triage Agent ● live
99.97%Success rate
312msAvg latency
4,208Runs today
Trusted by engineering teams at
Platform

Everything between a prompt and production.

Axis sits underneath your agent code, handling the parts that turn a working demo into something you can put your name on.

Orchestration

Compose multi-step agent workflows with typed handoffs, retries and parallel tool calls — no bespoke state machine required.

Evals

Run every model or prompt change against a versioned suite of real transcripts before it reaches production traffic.

Guardrails

Enforce input/output policies, PII redaction and tool-call allowlists at the runtime level, not scattered across prompts.

Observability

Full trace of every reasoning step, tool call and token — searchable, and attached to the support ticket it caused.

Guardrailed autonomy

Set spend, rate and scope limits per agent so autonomy has a ceiling your on-call engineer already agreed to.

Integrations

Native connectors for the model providers, vector stores and internal tools your stack already runs on.

Developer experience

Instrument an agent in four lines.

The SDK wraps your existing agent code — LangChain, a custom loop, or raw API calls — without a rewrite.

agent.ts
import { Axis } from '@axis/sdk';

// wrap your existing agent — no rewrite required
const runtime = new Axis({ project: 'support-triage' });

const agent = runtime.wrap(supportAgent, {
  evals: 'triage-suite-v3',
  guardrails: ['pii-redact', 'tool-allowlist'],
});

const result = await agent.run(ticket);
0% Runtime uptime SLA
0ms Added orchestration latency (p50)
0B Agent steps traced monthly
0+ Engineering teams in production
Pricing

Start free. Scale when it matters.

Every plan includes the full runtime — evals, guardrails and tracing aren’t a higher tier.

Starter

For solo builders and early prototypes.

$0/mo
Free forever
  • Up to 3 agents
  • 10k traced steps / mo
  • Community support
  • 7-day trace retention
Start free

Enterprise

For organizations with compliance requirements.

Custom
annual contract
  • Dedicated runtime
  • SSO + audit log
  • Custom data retention
  • Solutions engineer
Talk to sales
Customers

What breaks first is what we hear about first.

We had three separate retry loops fighting each other before Axis. Now it’s one runtime and one place to look when something’s wrong.

P
Priya Nair
Staff Engineer, Hearth

The eval suite caught a regression from a model upgrade two days before it would have shipped. That alone paid for a year of the contract.

M
Marcus Ade
Eng Lead, Northwind

Guardrails let us say yes to an agent-facing feature we’d been stalling on for months because nobody wanted to own the blast radius.

S
Sana Liu
VP Engineering, Meridian
Engineering blog

Notes from building agent infrastructure.

Give your agents a runtime, not a prayer.

Set up your first traced, guardrailed agent in under fifteen minutes.