Surfil/Products/Bench

// surfil bench · consumerpartly built

Leaderboard and a signed savings receipt.

Teams want to know if their agent setup is actually efficient, and they want proof they can hand to someone else - not a vibe.

Add Bench to your agents See pricing →

surfil · bench - acme/api

Quality on your repo

Opus - 42% fewer fixes / taskrecommended

ⓘ Bench is partly built. Some of what follows ships on a linked device today and the rest does not. Your dashboard marks each capability, and nothing shows a number it has not measured.

What it does

A receipt, not a vibe

Bench compares your measured savings against your own history and issues a signed, offline-verifiable receipt for the result - the same signature model used across every paid Surfil output.

✓Signed receipt - Every result is Ed25519-signed and independently verifiable, online or offline.

✓Self-referential leaderboard - Tracks your measured performance over time - no fabricated peer comparisons.

✓Consume → sign → verify - A receipt is generated at the moment savings are consumed, not estimated after the fact.

✓Exportable - Receipts travel with your Weave export - portable proof, not a locked-in dashboard number.

Proof, not promises

Bench produces one metric type on the shared spine: signed benchmark receipts. Every paid output is signed (Ed25519) and verifiable offline with no account. Zero-trace: your source never leaves the device.

$ surfil verify rcpt_8f2a
✓ VALID (offline · epoch 7)

In your terminal

Bench, as you'd actually see it

~/acme/api - surfil bench

$ surfil bench run --repo acme/api

# comparing this week against your measured baseline…

✓ tokens/task: −38% vs. baseline (measured)

✓ receipt issued: rcpt_8f2a · signed · epoch 7

$ surfil verify rcpt_8f2a

✓ VALID (offline · no account)

How it works

How Bench does it

1Bench replays your own measured history - dated runs, no synthetic peers.

2Each run compares tokens-per-task, cost and quality against your baseline.

3The result is consumed as a Credit and signed at that moment - consume → sign.

4The receipt verifies offline with surfil verify; no account, no dashboard trust.

5Receipts export with your Weave, so the proof travels with you.

Use cases

Where Bench earns its place

Model choice per repo

Find which model is cheapest-per-fix on your codebase, not in a vendor benchmark.

Proof for the budget owner

Hand finance a signed savings receipt instead of a screenshot they can't check.

Tracking drift

See whether your setup is getting more or less efficient over time, measured.

Claude CodeCursorCodexCopilotMCP agents

FAQ

Questions developers ask first

Is the leaderboard against other companies?

No - it runs against your own history only. We never fabricate peer comparisons.

What exactly is signed?

The measured facts of the run - token counts, dates, deltas - signed with Ed25519.

Can someone else verify my receipt?

Yes. surfil verify works offline for anyone holding the receipt - that's the point.

Works well with

Add Bench to your agents.

Core installs with Starter; Bench plugs into the same interception point - no second layer, no new setup.

Add Bench to your agents View pricing →

Leaderboard and a signed savings receipt.

A receipt, not a vibe

Proof, not promises

Bench, as you'd actually see it

How Bench does it

Where Bench earns its place

Model choice per repo

Proof for the budget owner

Tracking drift

Questions developers ask first

Is the leaderboard against other companies?

What exactly is signed?

Can someone else verify my receipt?

One spine - products compound

Cap

Proof

Core

Add Bench to your agents.