Features

See exactly what Spirehawk does.

Point your agents at one gateway and every session becomes something you can search, cost, and act on. Scroll the tour — the panel is the real interface, not a mockup.

Connect an agent

Route an agent through Spirehawk in one click.

Spirehawk detects the coding agents already on your machine and reroutes their traffic through the local gateway — so every run is observed. Original configs are backed up; disconnect any time.

  • Auto-detects installed agents — no restart
  • Traffic redirected locally, never to a cloud relay
  • Reversible — configs restored on disconnect

Connect your agents

Connect one agent to get started.

Connect all agents

3 detected · 1 connected

Claude Code

Connected

Detected · 2.1.3

Codex

Detected · 1.8.0

Cursor

Detected · 1.7.0

Conversations

Every session, as an inbox you can open.

Each agent run becomes a conversation with a cost, a model, and a machine-assessed state — laid out as a board of what needs you, what's active, and what's done.

  • Needs-you · Active · Done at a glance
  • Full-text search across transcripts
  • Open a run for its artifacts and tool rows

Waiting on you

2

Active

2

Done

2
Dashboard

Cost, activity, and what needs you.

One screen answers the three questions that matter when agents run all day — what it's costing, what's happening now, and what's blocked on you.

  • Approximate cost, honestly labeled
  • An attention inbox, oldest-waiting first
  • Your Day, as a run / wait / done timeline
Today in America/Los_Angeles
Updated 2m ago
Approx. cost
$18.40

↑ 12.4% vs yesterday

Cumulative known approximate cost today 9:12 AM to 1:58 PM: latest $18.40, range $2.00–$18.40
Tokens
4.2M

1,820 tokens completed/sec

Tokens completed by minute today 9:12 AM to 1:58 PM: latest 90K, range 30K–90K
Running
2

3 active conversations

Running conversation count over time 9:12 AM to 1:58 PM: latest 2, range 1–2
Needs you
2

1 awaiting instruction

Conversations needing you over time 9:12 AM to 1:58 PM: latest 2, range 0–2

Your day

3 conversations
Agent Rail

An ambient strip at the edge of your screen.

The Rail floats always-on-top and shows every running agent as a status glyph — so you never open the app just to check. Amber, the only color, means a human is needed.

  • A glyph per agent, always on top
  • Amber, the only color, means a human is needed
  • Hover to peek, click to jump to the session
SkillsComing soon

Reusable playbooks, installed into your agents.

A skill is a named prompt an agent loads on demand. Browse the library and deploy one into the specific agents that should have it.

  • Discover, My Skills, and Workspace libraries
  • Install into the agents you choose
  • Bundle a stack, or publish what already works

Skills

Create, share, and browse reusable prompts your agents can run.

Code review

Reviews a diff for correctness, edge cases, and risky changes.

Installed for agents

Write a PRD

Drafts a crisp product requirements doc from a short problem statement.

Product Team

Give feedback

Turns a vague concern into feedback that's specific, kind, and clear.

Marcus Chen

Security review

Maps trust boundaries and the likely attacks for a new feature.

Security Team
VaultComing soon

Give agents a placeholder, never the real key.

The Vault stores your secrets once, then mints short-lived, scoped placeholders for agents to use in their place — swapped for the real value at call time.

  • Store the real secret once
  • Mint scoped, time-boxed placeholders
  • Revoke one instantly — without rotating the secret

Vault

GitHub PAT

Personal access token · active

Connected 7 days ago

Live 2
Expired 1
Revoked 0

Placeholders

Task-scoped · claude-code
Live
ph_github_live_demo_a1b2c3

Minted 2h ago · expires in 1h · used 4×

Manual · claude-code
Live
ph_github_live_demo_f4e5d6

Minted 40m ago · expires in 20m · used 1×

Multiply what your agents deliver

Put Spirehawk in the path of your agents today — and turn what your team is already spending into productivity you can see, prove, and grow