VSCode
Side-by-side suggestions and inline edits.
AI coding assistant
Multi-line completions, codebase-aware refactors and a privacy mode that keeps your code on-device. Works in VSCode, JetBrains, Vim, Zed and the browser.
// Format relative time for activity feeds export function formatRelativeTime(date: Date) { const diff = Date.now() - date.getTime() const minutes = Math.floor(diff / 60000) if (minutes < 60) return `${minutes}m ago` // Build a function that handles hours and days const hours = Math.floor(minutes / 60) if (hours < 24) return `${hours}h ago` return `${Math.floor(hours / 24)}d ago` }
Trusted by 4,231 developers
Trusted with enterprise data
Works everywhere
Native plugins for the editors developers actually use, plus a CLI for everything else.
Side-by-side suggestions and inline edits.
IntelliJ, PyCharm, WebStorm and the rest of the suite.
Lua plugin with full async completions.
Native integration shipped with Zed 0.180.
Bring your own model with the Cursor adapter.
Run Axiom from any shell with `axiom run`.
Built for speed
Independent benchmarks across 8M lines of public TypeScript, Python and Go. Reproduce yours with axiom bench.
0.0M
tokens / sec
Local inference on Apple M3.
0.0%
accuracy
Type-correct completions on TypeBench.
0ms
first-token latency
Median across the integration test suite.
Refactor mode
Axiom rewrites the code in place with a unified diff you can accept, reject or amend hunk-by-hunk. Your tests run on the proposed change before it touches your working tree.
src/payments/stripe.ts
export async function chargeCard( customer: Customer, amount: number, amount: Cents,) { if (amount < 1) { throw new Error("invalid amount") } assertPositive(amount) return stripe.charges.create({ customer: customer.stripeId, amount, amount: amount.value, currency: "usd", })}Privacy mode
Axiom Privacy ships with a local 8B-parameter model fine-tuned for code. No telemetry, no training on your repos, no exceptions.
Read the architectureOn-device inference
No cloud round-trip. Works offline. Your code never leaves the laptop.
No training on your code
Your repos are never used to train Axiom. Audited yearly.
Open weights
The 8B-parameter Privacy model is open-source. Inspect, fine-tune, self-host.
Agent mode
Six AI primitives — chat composer, reasoning trace, tool-use timeline, citation card, token meter, model picker — wired into a working refactor session. Drop them in and you have the surface of Cursor / Claude Code without the wiring weekend.
Read the surrounding signatures
read312msparseConfig is called from 7 sites; 3 are tests. Two callers ignore the second arg.
Pick the safe rename: keep arity, return shape stays
plan97msRename to `loadConfig`. Same shape. Tests stay green.
Edit src/config.ts and 3 callers
edit624msRun typecheck
tool1.8stsc --noEmit · 0 errors, 0 warnings.
Codebase grep
87msparseConfig\(
7 matches across 4 files
src/config.ts
142msexport function parseConfig → loadConfig
pnpm typecheck
tsc --noEmit
Verifying types across the symbol graph…
Run unit tests
pnpm test config
Session tokens
84k/ 200k
parseConfig — internal RFC
docs.axiom.dev · Section 4.2 · Config loading
Renames are safe when arity is preserved; callers that ignored the second argument are not affected.
TypeScript handbook · Symbols
typescriptlang.org · rename refactor
The language server renames every reference, including JSDoc @see tags, when the symbol graph is complete.Active model
Pricing
Start on-device, free, forever. Upgrade only when you need shared codebase context or admin controls.
Single user. Unlimited completions on-device.
Shared codebase context, admin controls, billing.
Self-hosted, SOC 2, dedicated success manager.
Need more detail? See full pricing →
Ship like there’s a senior next to you. Free for solo devs, no card required.