# t2000

> The agent stack on Sui — wallet, payments, identity, and private AI for agents. CLI, SDK, and MCP.

## Docs

- [Agent ID](https://developers.t2000.ai/agent-id.md): A portable, on-chain identity for every agent on Sui — address-anchored, with a human-readable handle, an owner, and a public profile in the agent directory.
- [Agent SDK](https://developers.t2000.ai/agent-sdk.md): @t2000/sdk — the TypeScript SDK for Agent Wallets on Sui. Send USDC + USDsui gasless, swap via Cetus Aggregator, and pay any API in USDC over x402.
- [Architecture](https://developers.t2000.ai/agent-stack.md): The t2000 agent stack — one identity, one wallet, one balance. Hold and move USDC, pay for intelligence per call, prove who you are.
- [Agent Wallet](https://developers.t2000.ai/agent-wallet.md): One CLI, one MCP server, one skill set — hold + send USDC gasless, swap, pay any API over x402, all self-custodial.
- [Anchor receipt](https://developers.t2000.ai/api-reference/anchor.md): Anchor a confidential receipt on Sui — emits a `ReceiptAnchored` event committing the response wire_hash + attested workload_id. Idempotent: re-anchoring returns the existing tx. Public. See /confidential-ai/verify.
- [Get attestation](https://developers.t2000.ai/api-reference/attestation.md): Confidential-upstream attestation for a model. Returns the server-verified summary (cached); add `nonce` (or `full=true`) for the full ACI report — the raw TDX quote, keyset, endorsement, provenance, and freshness — bound to your nonce, so you can DCAP-verify the quote yourself and prove freshness.…
- [Chat completions](https://developers.t2000.ai/api-reference/chat-completions.md): Create a chat completion. Requests are served through the attested gateway, which retains no request bodies. A `phala/*` (confidential) response additionally runs on a verified upstream enclave — confirmed per response from the receipt's `upstream.verified` event — and returns an `x-receipt-id` head…
- [List models](https://developers.t2000.ai/api-reference/models.md): List the live, public model catalog — per-model pricing, context window, and privacy tier (`private` or `confidential`). No key required.
- [API reference](https://developers.t2000.ai/api-reference/overview.md): Base URL, authentication, endpoints, response headers, and errors for t2000 Private Inference (OpenAI-compatible).
- [Get receipt](https://developers.t2000.ai/api-reference/receipts.md): The signed per-response receipt, binding your request/response hashes (not bodies) to the attested workload. `{id}` is the `x-receipt-id` from a confidential completion. Public. See /confidential-ai/receipts.
- [Get session](https://developers.t2000.ai/api-reference/sessions.md): The attested-session record a receipt references (`upstream.verified.session_id`): the verified upstream channel, its binding, typed TCB claims, and evidence. Public. See /confidential-ai/attested-sessions.
- [Authentication](https://developers.t2000.ai/authentication.md): Three ways to authenticate to t2000: a Private Inference key, your non-custodial wallet, or keyless x402 pay-per-call — all on one Passport identity and balance.
- [Changelog](https://developers.t2000.ai/changelog.md): What's new in the t2000 stack — the notable changes, newest first. Every version's full notes live on GitHub Releases.
- [CLI Command Reference](https://developers.t2000.ai/cli-reference.md): Every t2 command — wallet, payments, identity, Private Inference, MCP. One page, generated against the live CLI.
- [Attestation](https://developers.t2000.ai/confidential-ai/attestation.md): The attestation report proves which TEE workload served your request and publishes the keys it signs with. Verify it first — every other guarantee depends on it.
- [Attested sessions & claims](https://developers.t2000.ai/confidential-ai/attested-sessions.md): The verified upstream channel behind a confidential response — its channel binding, typed TCB claims, and evidence — resolvable from a receipt's session_id.
- [FAQ](https://developers.t2000.ai/confidential-ai/faqs.md): Common questions about t2000's Confidential tier — attestation, receipts, the Sui anchor, the trust boundary, and how it differs from the Private (ZDR) tier.
- [How Confidential AI works](https://developers.t2000.ai/confidential-ai/how-it-works.md): The t2000 Confidential tier serves inference from a verified GPU-TEE and proves each response with an attestation report, a signed receipt, and a Sui anchor you can check yourself.
- [Providers](https://developers.t2000.ai/confidential-ai/providers.md): The verified upstream providers behind t2000's confidential models, the channel binding each enforces, and what each attests — the receipt names who served a given request.
- [Receipts](https://developers.t2000.ai/confidential-ai/receipts.md): A signed receipt is a per-response transparency log that binds your request and response hashes to the attested gateway — hashes, never bodies.
- [Trust boundary](https://developers.t2000.ai/confidential-ai/trust-boundary.md): What the Confidential tier proves, what it still trusts, and where your plaintext is visible — stated plainly. A wrong 'verifiable' claim is worse than honest ZDR.
- [Verify a response](https://developers.t2000.ai/confidential-ai/verify.md): Check — don't trust — a confidential response by its receipt id. t2 verify reads the on-chain Sui anchor and recovers the receipt signature, both trustlessly, and fails closed on any mismatch.
- [Agent Console](https://developers.t2000.ai/console.md): The no-code path — sign in with Google (zkLogin) to manage API keys, billing, usage, and your agents. Same Passport + balance as Audric.
- [create-t2-app](https://developers.t2000.ai/create-t2-app.md): Start a router-wired agent project in one command — npm create t2-app@latest.
- [Introduction](https://developers.t2000.ai/index.md): The agent stack on Sui — wallet, payments, identity, and private AI for agents. CLI, SDK, and MCP.
- [Models](https://developers.t2000.ai/models.md): The t2000 model catalog — open, frontier, and confidential — behind one key, billed per token from one balance. Curated, not a 250-model dump.
- [Pay any API](https://developers.t2000.ai/pay-any-api.md): Pay any API in USDC over x402 — per call, gasless, no signup, no API keys.
- [Private Inference](https://developers.t2000.ai/private-inference.md): api.t2000.ai — an OpenAI-compatible inference API. Private by default (zero data retention); verifiably confidential (GPU-TEE + Sui-anchored signed receipts) when it matters. Every model behind one key, pay-as-you-go.
- [Quickstart](https://developers.t2000.ai/quickstart.md): Install, fund a wallet, pay APIs per call, register an identity, and run private models — five steps, each one command.
- [Make AI music](https://developers.t2000.ai/recipes/ai-music.md): One sentence in, an original track out. Claude crafts the prompt, Stable Audio 2.5 renders real, commercial-grade music — paid as you go in USDC.
- [What's new in AI this week](https://developers.t2000.ai/recipes/ai-research-digest.md): A weekly digest of the AI research that actually matters — specific papers with arXiv IDs, plus the debates and launches to track. Exa + Perplexity find it, Claude distills it. Paid as you go.
- [Tear down a competitor](https://developers.t2000.ai/recipes/competitor-teardown.md): A domain in, a positioning teardown out. Firecrawl maps the site and pulls structured facts, Claude synthesizes the angle. Live data, paid as you go, no API keys.
- [Concept → demo asset](https://developers.t2000.ai/recipes/concept-asset.md): From idea to image + pitch + voiceover in one prompt. Chains fal.ai + Anthropic + ElevenLabs over x402.
- [Deep-dive a crypto asset](https://developers.t2000.ai/recipes/crypto-research.md): One token, the full picture. Market data + headlines + a web-grounded narrative with citations — CoinGecko + NewsAPI + Perplexity, paid as you go.
- [Inspect a sketchy link](https://developers.t2000.ai/recipes/link-inspector.md): A URL in, a safety verdict out. VirusTotal scans it across 70+ engines, Jina reads what's actually on the page, ScreenshotOne shows it — Claude gives the verdict. Live data, paid as you go, no API keys.
- [Localize an audio clip](https://developers.t2000.ai/recipes/localize-clip.md): An English clip in, the same clip in another language out. fal Whisper transcribes, DeepL translates, ElevenLabs re-voices it. A full audio round-trip, paid as you go, no API keys.
- [Mail a physical card](https://developers.t2000.ai/recipes/mail-card.md): Agent writes the message, generates the front, and ships a real postcard via USPS. ~$2.08 end to end.
- [Prep for a meeting](https://developers.t2000.ai/recipes/meeting-prep.md): A name and a company in, a tight pre-meeting brief out. Perplexity + Exa + Serper pull the company, the person, and what's recent — Claude flags what matters. Paid as you go, no API keys.
- [Run a model bake-off](https://developers.t2000.ai/recipes/model-bakeoff.md): One prompt in, five models' answers out — side by side. Claude, GPT, Gemini, Groq (Llama), and DeepSeek all answer the same prompt, then Claude judges. One wallet, no five API keys.
- [Morning market brief](https://developers.t2000.ai/recipes/morning-brief.md): Chain CoinGecko + NewsAPI + Anthropic over x402. Three paid calls, ~$0.06, zero API keys.
- [Research with citations](https://developers.t2000.ai/recipes/research-citations.md): A question in, a sourced answer out. Exa finds the primary sources, Perplexity grounds the answer in live web citations — paid as you go, no scraping setup.
- [Research a stock](https://developers.t2000.ai/recipes/stock-research.md): Ticker in, a sourced equity brief out. Real-time quote, trend, and the news moving it — Alpha Vantage + NewsAPI synthesized by Claude, paid as you go.
- [Make an illustrated storybook](https://developers.t2000.ai/recipes/storybook.md): One sentence in, a printable picture book out. Claude writes the story, Flux illustrates every page in a consistent style, and PDFShift binds it server-side. Real generative art a code sandbox can't fake. ~$0.34.
- [Transcribe & summarize audio](https://developers.t2000.ai/recipes/transcribe-summarize.md): A recording in, key points out. fal.ai Whisper transcribes any public audio URL, Claude pulls the highlights — paid as you go, no upload.
- [Plan a trip](https://developers.t2000.ai/recipes/trip-planner.md): A city and dates in, a planned trip out. SerpAPI finds the flights, Google Maps routes the days, OpenWeather checks the forecast — Claude turns it into an itinerary. Paid as you go, no API keys.
- [Write code, then run it](https://developers.t2000.ai/recipes/write-and-run.md): Agent writes a script with Claude, then executes it via Judge0 to verify the output. Two paid calls, ~$0.04.
- [Sell your API on Sui](https://developers.t2000.ai/sell-your-api.md): Accept per-call USDC from any Sui agent — no accounts, no API keys for buyers, instant settlement.
- [t2 code](https://developers.t2000.ai/t2-code.md): The private coding agent you can verify — a multi-agent coding CLI running on t2000 Private Inference, open models first, GPU-TEE confidential on demand.
- [Use from your agent](https://developers.t2000.ai/use-from-your-agent.md): Wire the t2000 Agent Wallet into the agent you already run — Audric, Claude Code, Cursor, or Codex. 3–4 steps per path, one paste per step.
- [OpenAI-compatible apps](https://developers.t2000.ai/use-with-your-tools.md): t2000 is OpenAI-compatible — point any OpenAI client at api.t2000.ai/v1 with your key and every model (private + confidential) works as a drop-in, billed from one balance.

## OpenAPI Specs

- [openapi](https://developers.t2000.ai/openapi.json)