Observe

See what your agents are really doing.

Every conversation, every request, every dollar — visible in one place. Spirehawk records each tool call, its arguments, and the decision that allowed it, so you can answer what happened, who approved it, and what it cost without stitching logs together by hand.

Today

The problem with black-box agents

An agent run today is reconstructed by hand from scattered tool logs that know nothing about the task or person or price behind them.

How it works

Observability from the gateway, not the agent.

Spirehawk sits between your agents and your tools. Because every request already passes through it, the record is complete by construction — nothing to install in the agent, nothing to forget to log.

01

Point your agents at the gateway

Change one base URL in the harness — Claude Code, Codex, Cursor. No SDK, no agent instrumentation.

02

Every request writes the ledger

Each tool call is logged as it passes: name, arguments, decision, result, timing, and cost — in order, as it happens.

03

Every entry carries its authority

The task that authorized the run and the human who approved it travel with every call.

04

Ask anything, any time

Replay a run, search a conversation, export the record. The trail is already there when you need it.

What you get

One live view of every run.

No scattered logs. No tab-checking ritual. Just a complete record of agent activity that is easy to search, replay, and export.

Every call, captured

Tool name, arguments, result status, and timing are logged at the gateway, the one place every request already passes through.

Full traceability

Each entry is written at the gateway as the call happens, so every run leaves a complete, ordered record.

Replayable & exportable

Step through any run after the fact, with the full call history in one place.

Conversations

Open the conversation behind any number.

Every run is a conversation you can step into — the full exchange, the tickets it served, the artifacts it produced, and what it cost, on one screen.

Per-conversation breakdown

Tokens, messages, cost, and model — totaled per conversation and tied to the tickets it worked.

Artifact trail

Every file, config, doc, and pull request the conversation produced, timestamped in order.

Search inside

Full-text search within any conversation — jump straight to the exchange that matters.

Outcome

“How many tokens did I spend?”
Every answer is a single lookup, not an investigation.

how many tokens did my agents spend today?
48.2M tokens · $2,639 metered today · ▲ 10.9% vs yesterday

See every action. Know every cost.

Put Spirehawk in the path of your agents and turn black-box runs into something you can inspect, measure, and trust.