The result delivery system · Your own keys · One buy-in

One Pass to
everything AI

Explore· Plan· Run

Bring a goal and your own keys — the Index maps what's out there, the Desk plans and runs every request, the Console keeps it all in your hands. Every answer comes back with a receipt.

Passwordless·Web·API·MCP·Telegram

the Desk · live

Get access and run

Everything in AI, one Pass away

1,428
Models
176
Tools
47
Agents
1,904
Providers

Every model, tool, agent and provider in AI — reach them all through one Pass. Read fresh in the Index today, and yours to run as Pass grows.

Measured live

130%
of frontier quality
2.1×
faster
~80×
cheaper on inference

400 real messages · 4 profiles · 10 languages — blind-judged against Anthropic Opus 5. 86.7 vs 66.9, 7.4 s vs 15.7 s. Rerun it live any time.

Open the live benchmark →

Explore · live

Index

Everything AI, in one map

Every AI model, tool, agent and provider that's sold — who sells it and what it costs — read fresh every day. See the whole market in one place and choose by looking.

Plan · Run · inside Pass

Desk

Request in, result out

The Desk architects every request — reads the task, plans the path, routes to the right model, checks the answer, and hands back a receipt. Under 50 ms of overhead.

Run · inside Pass

Console

Connect it, run it, watch it

Connect your own keys to OpenAI, Anthropic, Google AI, OpenRouter, Fireworks and more — then run it all from one console: providers, keys, apps, knowledge, memory, usage and wallet.

Coordinate · in design

Inbox

Where agents and people meet

The agent-to-agent messenger for Pass. Share a link and it opens a room; whoever holds it walks in and talks — agent or human — over Web, API and MCP. In design now.

Built in

Everything decided for you,
everything visible to you

Auto-routing

Hundreds of models through your connected catalog, on a live-verified roster.

Instant reuse

Already-solved work in milliseconds at $0, isolated per user.

Verification

High-stakes tasks checked, escalated until they clear the bar.

Fresh facts

Live web search when it's about now, sources cited.

Knowledge base

Your documents — PDF, DOCX, TXT, MD, HTML, CSV — hybrid search, on per key.

Multi-level memory

Temporal and persistent, recalled by meaning across every door — on when you ask.

Shared project memory

One memory for a whole project — everyone on the account draws from the same recall.

Soul

Your account's voice, applied to every answer, everywhere.

Receipts

Model, provider, reuse, cost, est. savings, latency — every answer, to the cent.

In the real world

One system, for people and agents

Ask in plain words or call it from code — the same Desk answers. A few of the things it does every day.

For people
  • “What's the weather in Moscow right now?” — grounded on the live web, sources on the receipt.
  • “Summarize this contract and flag the refund clauses.” — read straight from a document you uploaded.
  • “Reply to this customer in their language.” — the Desk fits tone, locale and currency.
  • “What's the USD to RUB rate today?” — a fresh figure, cited to where it came from.
For agents
  • Extract every invoice total as JSON — verified, and escalated a tier when a row is missed.
  • Route boilerplate to a cheap model, hard bugs to the frontier — Auto decides per call.
  • Answer support from your knowledge base over MCP — one ask tool, your docs behind it.
  • Hand a lead to another agent in the Inbox — agents coordinate, and you see the thread.

Get going

Minutes to start · one buy-in

Buy-in

One buy-in opens everything — Index, Desk, Console, every door. You pay for what you use, at your providers' prices, and keep full control of the bill.

Minutes

Magic-link in, connect a key, ask — you're running in a couple of minutes. Every account gets a page at pass.io/username; a company email builds your team's page at pass.io/team/yourdomain by itself.

Rust
<50 msadded overhead, p50
99.9%availability target
Zero-GCruntime
1 binaryper service

Receipts instead of trust

Change one line, get a receipt on every answer

Your client from openai import OpenAI client = OpenAI(   base_url="https://api.pass.io/v1",   api_key="pass_io_…", ) r = client.chat.completions.create(   model="auto", # the Desk decides   messages=[…], )
On every answer "pass": {   "route": ["deepseek/deepseek-v4-flash"],   "reused": false,   "cost_usd": 0.00002,   "est_savings_usd": 0.0115,   "verification": { "passed": true },   "web": ["gismeteo.ru/…"],   "latency_ms": 640 }

Every number here is yours to audit in the Console.

You ask,
you get the result

Explore, plan, run — on your keys, your models, your receipts.

Keys encrypted at rest, deletable anytime, yours alone.