Lattice / Developer API Closed beta

Your tools.
One developer API.

Connect your apps and coding tools to available models through an OpenAI-compatible endpoint. Account-scoped keys, clear usage, and an exact model choice.

Join the closed beta Read API docs

API access is invite-only. Have a beta access code? Create your account to begin. Paid access depends on current availability in Billing. Browser chat is reserved for admins.

GET /v1/modelsBearer authorization

Start with the catalog.

Copy your Base URL from Account, including /v1. Discover your available model IDs before making a generation request.

bash

curl "$LATTICE_BASE_URL/models" \
  --header "Authorization: Bearer $LATTICE_API_KEY"

Example only. This page does not send a request. A key does not by itself grant model access.

A short path to your first request.

Create a credential

Name a key for your trusted tool. Save the secret when it is displayed, then keep it on your server or local client.

Choose your model

Discover the available catalog with your key. For Go, use the exact qualified model shown in Billing.

Know your allowance

Review rolling usage and current access in Billing. Included allowance and purchased PAYG cash stay separate.

Planned launch offer

Go API · $10/month

$40 in virtual reference usage per rolling 30 days, with shorter rolling limits. These credits measure API use; they are separate from purchased cash.

Purchases pausedCheck Billing and availability

Rolling 5 hours: $4virtual usage
Rolling 7 days: $20virtual usage
Rolling 30 days: $40virtual usage

One premium request at a time across all your API keys. An additional 8M generated-token guard applies per rolling 30 days.

Separate PAYG cash, explicit opt-in. Included requests never fall back to cash. Automatic top-ups are off.

Reference usage weights: $0.15 fresh input / $0.03 verified cached input / $0.50 generated output per million tokens. Limits are configurable; Billing shows your account's current policy. Paying an invoice does not reset rolling usage windows.

GLM-5.3-Flash, 262,144-token context and one full project retained for seven days are launch targets awaiting qualification. They are not current availability or cache guarantees. Purchases require serving, payments and cache qualification.

POST /v1/chat/completions

Use the client you already know.

Set your Base URL, saved key, and exact model ID. This Bash example requires curl and jq, and stops if the chosen model is unavailable; no model is substituted silently.

Function tools depend on the selected model. Vision, audio, embeddings and external provider overflow are not included in the Go launch offer.

Setup and error recovery

bash

# Requires curl and jq. Copy Base URL from Account → API keys.
# LATTICE_BASE_URL includes /v1. Set your key and exact model ID.
set -euo pipefail
: "$LATTICE_BASE_URL" "$LATTICE_API_KEY" "$LATTICE_MODEL_ID"
curl --fail --silent --show-error "$LATTICE_BASE_URL/models" \
  -H "Authorization: Bearer $LATTICE_API_KEY" |
  jq -e --arg model "$LATTICE_MODEL_ID" '.data | any(.id == $model)' >/dev/null
curl --fail --show-error "$LATTICE_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $LATTICE_API_KEY" \
  -H "Content-Type: application/json" \
  --data "$(jq -nc --arg model "$LATTICE_MODEL_ID" \
    '{model:$model,messages:[{role:"user",content:"Hello"}]}')"

Default requests use included allowance. PAYG requires separate spending permission and an explicit billing-mode header.

Clear data handling.

API clients own their conversation context. The selected inference host processes requests in plaintext. Usage metadata helps operate the service. Read the privacy notice for retention and processing boundaries.

Review your request history.

Account → Usage shows recorded activity. Billing shows verified allowance and spending state.

Open request history