Surfil

What is Surfil?

Surfil is an on-device control plane for AI coding agents such as Claude Code, Cursor, Codex, and Copilot. It installs with one terminal command, watches agent traffic on your machine, measures real token spend, blocks requests carrying keys or injected instructions, and keeps shared memory your agents can reuse across sessions.

Top Features:

• Cap cost ledger: shows each request's cost from the provider's own token counts.
• Guard screening: stops requests carrying API keys, tokens, or hidden prompt injections.
• Mind memory: keeps encrypted facts agents learn and flags them when code changes.

Use Cases:

• Cost audits: see where agent token spend goes before the monthly bill arrives.
• Secret protection: catch credentials in agent payloads before they leave the laptop.
• Shared context: let Claude Code and Cursor reuse past decisions about your codebase.

Who Can Use Surfil?

• Solo developers: see what coding agents spend and catch risky tool calls.
• Engineering managers: compare agent costs across a team with shared dashboards.
• Security teams: enforce policies on agent traffic without sending code off device.

Pricing

• Starter ($5 per month): core install, Guard monitoring, and the basic Cap cost ledger.
• Pro ($25 per month): full Cap, Mind, Radar, Bench, and enforced Guard after a $1 trial.
• Team ($20 per seat): everything in Pro plus team memory, dashboards, and pooled credits.

Pros and Cons

Pros:

• Code stays local: source never leaves the machine and telemetry is only metadata.
• Verifiable numbers: every paid output is signed and can be checked offline.
• Broad agent support: one install covers Claude Code, Cursor, Codex, Copilot, and more.

Cons:

• No free tier: plans start at five dollars monthly with only a paid trial.
• Limited platforms: installation needs a Mac or Linux computer and a terminal.
• Savings not automatic: the cost tool measures spend today but does not cut it.

FAQs:

1) Which coding agents does it work with?

Claude Code, Cursor, Codex, Copilot, and over twenty more work from one install.

2) Does my source code leave my machine?

No, heavy work runs on device and only metadata telemetry leaves it.

3) Will it lower my token bill?

It measures real spend now, and automatic cost reduction is planned next.

4) What is a credit?

One credit equals one signed output, such as a savings receipt or benchmark.

5) Can I uninstall it cleanly?

Yes, uninstalling restores every agent config exactly and leaves no residue behind.