翻訳待ち：Surfil On-device control plane for AI coding agents

AI サービスが一時的に利用できないため、復旧後に翻訳を補完します。

// on-device control plane for AI coding agents One install. Cheaper, safer, and remembered. Surfil runs beside Claude Code, Cursor, Codex & Copilot. The tools that route through it get cheaper and safer; editors like Cursor connect through MCP - every claim proven with a signature, and your source never leaves the machine. Install SurfilSee pricing →How it works → surfil · your control plane ▤ Saved / mo $412 ⛨ Blocked 27 ≋ Memory 94% Savings trendsigned ✓rcpt_8f2aepoch 7 · offlineVALID !AWS key in payloadacme/apiblocked Works with Claude CodeCursorCodexCopilotMCP agents byte-exact never busts your prompt cache 0 lines of source leave device 100% offline-verifiable receipts 1 interception point Proof you can verify yourself Don't trust our numbers. Check the signature. Every Surfil claim ships as an Ed25519-signed receipt. Verification runs offline, with no account - on your machine, against a public key. This is what it looks like: ~/acme/api - surfil $ surfil consolidate # scanning 6 MCP servers across Claude Code + Cursor… ✓ consolidated 6 → 1 endpoint saved ~48k tokens/op $ surfil audit ✓ recoverable spend this week: $96.40 (signed: rcpt_8f2a) $ surfil verify rcpt_8f2a ✓ VALID (offline · epoch 7 · no account) ⓘ The numbers above are illustrative. Yours are measured on your own repos. We publish no fixed savings percentage, because the honest answer depends on your traffic: what Surfil guarantees is that every figure is measured and signed, and that interception is byte-exact, so the provider prompt cache that already removes most of your input cost keeps hitting. The problem Agents are expensive, risky, and forgetful Not because they're bad - because nothing sits beside them watching cost, safety and memory across all of them at once. Expensive Agents re-send the same context and re-read the same files. Token spend creeps up and nobody can point to why - or prove what's recoverable. Risky Agents run shell commands, hit the network, and paste payloads. One leaked key or injected tool-call is discovered after it already happened. Forgetful Every session starts from zero. Agents re-derive your architecture, re-ask settled questions, and forget the decisions you already made. What Surfil does Cheaper, safer, remembered - on one spine Three products carry the consumer story. All of them are stages on the same pipeline, writing to the same signed knowledge network. Cheaper Cap measures where your input cost actually sits: the 1.25x premium on every cached write, and how often it is read back. Every figure is the provider’s own count, measured before and after. Details → Safer Guard blocks secrets and OWASP-scored injection pre-flight, on-device. Simulate before you enforce; nothing runs unchecked. Details → Remembered Mind gives you memory that follows you across agents - facts carry provenance, stale is auto-flagged, and everything is portable. Details → How it works Install. Intercept once. Verify the receipt. Three steps between your current agent bill and a signed number you can hand to anyone. 1 · Install One install beside Claude Code, Cursor, Codex and Copilot. Core consolidates your scattered agent configs and bootstraps Weave. 2 · Intercept once A single on-device interception point, byte-exact passthrough. Every product is a stage on the same pipeline; no second proxy, ever. 3 · Verify the receipt Savings and outputs are measured, then Ed25519-signed. Run surfil verify on any receipt - offline, no account. The signature is the proof. The catalog Eleven products, one interception point Each product produces exactly one metric type on the shared spine - so every number has a source and a signature. Core Consumer The install that bootstraps everything else. Cap Consumer Cut token spend with measured, signed savings. Guard Consumer Sandbox and block risky tool calls before they run. Mind Consumer Local, encrypted memory your agents can reuse. Radar Consumer Observability and rewind - metadata only. Bench Consumer Leaderboard and a signed savings receipt. See all 11 products → Why you can trust it Four rules we can't break These aren't policies - they're architecture. Breaking any of them would require rebuilding the product. Zero-trace Source never leaves the device. Telemetry is metadata only; memory syncs as ciphertext. Signed outputs Every paid output is Ed25519-signed and verifiable offline - no trust-us dashboards. Honest claims Measured facts, never “certified”. Savings reported in tokens, at the conservative floor. One layer A single interception point, by rule. We never chain a second proxy to sell you more. Read the security architecture → Pricing Execution is unmetered. Signed output is metered. You're never charged for running your agents - only for the server-signed value-outputs Surfil produces. One signed output = one Credit. Starter $5/mo Core install, Guard monitor and the Cap cost ledger Choose Starter recommended Pro $25/mo Guard enforce, Cap full, Mind, Radar · Bench · Pilot Start with Pro Team $20/seat Everything in Pro, plus team dashboards and pooled credits Talk to sales Enterprise custom Proof · Fleet Control · Charter · Policy + SSO/SLA Contact sales FAQ Questions developers ask first Does my source code leave my machine? No. Zero-trace is architectural, not a setting. The heavy work runs in the on-device runtime; telemetry is metadata only and memory syncs as ciphertext. How much will I actually save? That depends on whether your agents already use prompt caching, and Surfil measures which case you are in. If they do, the provider cache is already removing most of your input cost, and byte-exact interception means Surfil never busts it. If they do not, Surfil can safely prune context an agent already superseded. We publish no fixed percentage: you get the measured figure on your own traffic, in tokens, never an invented dollar amount. Which agents does it work with? Claude Code, Cursor, Codex, Copilot and 30+ more, from one install. It connects each the way it can - full traffic interception for CLIs and MCP, MCP for editors like Cursor. Run surfil doctor to see how each connects. What is a Credit? Execution is unmetered. One server-signed value-output equals one Credit. You're metered on signed output, never on running the agent. Can I verify the savings myself? Yes. Run surfil verify on any receipt - offline, no account. The signature (Ed25519) is the proof, not a dashboard. What happens if I uninstall? A clean uninstall restores every agent config byte-for-byte and leaves zero residue. Export your Weave anytime with surfil export. Install once. Verify everything after. Core installs beside your agents. Cap, Guard and Mind light up for the tools that route through Surfil - nothing leaves your device, and uninstalling restores every config byte-for-byte. Install SurfilTalk to sales