Get started
Tokentide documentation
Frontier and open-weight models through an OpenAI-compatible API, at a live discount to API list prices. Zero data retention, guaranteed on Private models.
Tokentide is a drop-in, OpenAI-compatible API for frontier and open-weight models, at a discount to API list prices. To switch, change two things: the base URL and the API key. Prices follow a published intraday tide table, and an opt-in Low Tide lane sells spare capacity even cheaper. Zero data retention, guaranteed on Private models.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.tokentide.app/v1", # was https://api.openai.com/v1 or https://api.venice.ai/api/v1
api_key=os.environ["TOKENTIDE_API_KEY"], # a Tokentide key, tt_live_…
)New here? Start with the Quickstart. Prefer to explore? Open the API reference and send real requests from the built-in console.
Base URL
Every request goes to one base URL: https://api.tokentide.app/v1. The API also answers on the Venice-style alias
https://api.tokentide.app/api/v1/*, so Venice SDKs only need a new host and key.
Tokentide runs in the United States (AWS US East). Everything it stores (account data, request metadata and the ledger) is kept in the United States as well, and request content is never stored. Venice decides where inference runs. See Privacy and ZDR.
Guides and reference
- QuickstartChange two settings, the base URL and the key, and make your first request with curl, the OpenAI SDKs, the Vercel AI SDK or LangChain.
- Lanes and Tide PricingHow the price moves through the UTC day, the opt-in Low Tide lane, every X-Tokentide-* header and the market API.
- Privacy and zero data retentionWhat we never store, the metadata we keep and for how long, privacy modes and model privacy tiers.
- End-to-end encryptionEncrypt prompts to a hardware-attested key fetched directly from Venice, so Tokentide only ever relays ciphertext.
- Large Payload LaneSend files above the 4.5 MB direct limit: client-side AES-256-GCM encryption and one-time private staging.
- Video and musicQueue video, music and sound-effect jobs, collect the files, the daily cutoff before 00:00 UTC, and automatic refunds of failed jobs.
- Claude connector and pluginUse Tokentide from Claude: the connector's tools and sign-in, what it leaves out, and the plugin with the viral product video skill.
- Billing and spending limitsPrepaid credit by card, auto top-up, daily, weekly and monthly spending caps, e-mail alerts, and usage and cost tracking.
- ErrorsThe error envelope, every Tokentide error code, upstream errors and when to retry.
- Rate limitsPer-key and per-account limits by trust tier, the rate-limit headers, and body, duration and stream limits.
- API referenceEvery endpoint, generated from the OpenAPI 3.1 document. Send real requests from the console.
- SDK helpersTypeScript and Python helpers for Tokentide-native endpoints, response metadata, the Large Payload Lane and the Tide Runner.
- ChangelogDated changes to the Tokentide API.
Endpoints at a glance
Voice cloning and the voice changer are not offered: they keep voice samples upstream for days. Venice account routes
(/api_keys, /billing) are never proxied.
Status
GET /v1/status (no authentication) reports the health of the API, with Tokentide and the upstream (Venice) shown
separately, and the state of the current epoch:
{
"object": "status",
"status": "operational",
"components": { "tokentide_api": "operational", "market": "operational", "upstream_venice": "operational" },
"market": {
"epoch": "2026-10-31",
"state": "open",
"tick_age_ms": 4200,
"sales_paused": false,
"lowtide_open": true
},
"open_breakers": 0,
"deployment": { "id": "dpl_…", "commit": "b755503…" },
"checked_in_ms": 12,
"at": "2026-10-31T18:00:00.000Z"
}status is operational, degraded or paused (sales paused). Incidents and maintenance are posted on
status.tokentide.app.