The Token Archive
Est. 2026 · Prompt economics

The Token
Archive

Pay less. Ship the same.

Save 30–60% of your AI input tokens with a one-line, OpenAI-compatible base URL switch. Input tokens only — not a guaranteed cut to your provider bill.

Get API keySee coverage
free tier · no card · keys prefixed ta_ 
How it works
01

Create a key

Mint a ta_ key in the dashboard.

02

Point your base URL

OPENAI_BASE_URL → the archive.

03

Read the savings

X-TA-* headers on every response.

What it works with

Anything that lets you point at a custom OpenAI base URL.

Cursor

Set Override OpenAI Base URL in Settings → Models. Chat and Composer route through the archive.

Cline / Roo Code

Choose the OpenAI Compatible provider, paste the base URL and your ta_ key.

Continue

Set apiBase in ~/.continue/config.json, or run npx token-archive init --target continue.

Hermes Agent

Set OPENAI_BASE_URL in ~/.hermes/.env — compresses Hermes → model traffic.

Aider

Add openai-api-base to .aider.conf.yml and use your ta_ key.

Open WebUI / LibreChat

Add a custom OpenAI connection with the archive URL and your ta_ key.

OpenAI SDKs + LangChain

Pass baseURL / base_url with your ta_ key. openai-python, openai-node, LangChain, LlamaIndex, curl.

Honest limits. The archive cannot route GitHub Copilot or Cursor Tab — those completions run on fixed, first-party endpoints with no base URL override. Everything else that speaks the OpenAI Chat Completions API is fair game.

Start with a key, watch the meter.

Sign in with your email, mint a ta_ key, and the dashboard shows tokens saved per request as they land.

Get API key