Latitude 10 Agent Trace Kit

What is the Latitude 10 Agent Trace Kit?

When an agent goes wrong, “the request failed” tells you nothing. The Trace Kit records each task as one session: every model call, tool call, HTTP call and its retries, payment and sub-agent, in order and across processes, with latency, tokens and dollars on every step, and an outcome at the end.

A TypeScript SDK you add to your agent (one command adds it for you), and a small viewer you run yourself on a SQLite file or Postgres. It speaks OpenTelemetry in and out, so it works next to what you already run. Code you own; runs on Node 22, in Docker or on Fly.io.

Want it installed for you? Ask for a fixed-price quote on the contact page.

You run this code and are responsible for it: review it, test it on staging, and check that it meets the laws, platform rules and security standards that apply to you before production. It is not legal, compliance or financial advice, and using it doesn’t make Latitude 10 your payment processor or service provider.

What do you get?

• One trace per session across processes: your agent, the Agent Gateway, MCP servers and paid APIs in one waterfall
• Cost that includes money: estimated token cost and settled x402, MPP and L402 payments, kept apart; unknown models flagged, never $0
• Retries folded under the step that made them; an outcome (ok, error, limit) on every session
• Anthropic SDK, Claude Agent SDK, OpenAI, OpenAI Agents SDK and Vercel AI SDK integrations, plus any OpenTelemetry instrumentation
• Private by default: content off until you turn it on, redacted twice when on, and expiring sooner than the rest
• Sessions hash-chained when sealed, so a verify command catches any edit or deletion
• A UI with filters, usage by model, tool and merchant, alerts, share links and export as a test fixture

Every kit includes

• The full source code, yours to change
• A setup wizard: node setup.mjs
• A versioned release with a changelog
• A download link that always serves the latest release

How do you set it up?

Boots the viewer and runs a demo agent through five sessions, without any API key: a happy path, an API that fails twice before it works, a tool behind an x402 paywall, a sub-agent in a second process that joins the same trace, and a failure that fires an alert. Then it seals the sessions, verifies the record and prints one as a waterfall.

You'll need: Node.js 22.13+. Optional: Postgres or Supabase, a Slack incoming webhook, an OpenTelemetry backend to forward to.