What is Endlesstokens?
Endlesstokens is a hosted API for people using coding agents, editors, scripts, and other AI clients that support Chat Completions, Responses, or Anthropic Messages. Instead of charging by token or request, it provides an uncapped usage window through a one-hour, one-day, or three-day pass. Users set the API base URL and key in a compatible client, then use hosted open-weight models or let et:auto choose a model for each conversation. Each key supports two parallel sessions and a 128K context window.
Endlesstokens Features:
Flat Time Pass: A pass lasts one hour, one day, or three days from payment, with no subscription or automatic renewal.
Uncapped Requests: There is no token or request cap during the pass, though each key is limited to two parallel sessions.
API Compatibility: The API supports Chat Completions, Responses, and Anthropic Messages for clients that let users set a base URL and key.
Coding Client Setup: Setup guides are available for coding tools including Aider, Claude Code, Cline, Codex, Kilo, OpenCode, and Roo.
Automatic Model Choice: The et:auto option selects a model for each conversation, and passes include access to that model-selection option.
Image And Context: Requests can include up to five images, and Qwen3.8-27B is served with vision, reasoning, tool calling, and a 128K context window.
Endlesstokens Benefits:
Predictable Pass Cost: Paying for a time window rather than metered usage makes it easier to plan costs while an agent runs.
Fewer Meter Checks: With no token or request counter to watch, users can focus on the task instead of tracking each agent call.
Flexible Client Choice: Compatibility with several API formats lets users keep working in supported clients without installing a dedicated plugin.
Prompt Privacy: Because prompts are not stored and only request metadata is retained, users do not have their prompt text kept by the service.
Room for Long Inputs: The 128K context window gives compatible workflows space to provide substantial conversation context to the model.
Use Cases:
Coding Agent Sessions: Use a pass to run coding-agent work in Claude Code, Cline, Aider, Codex, or another supported client.
Long Agent Loops: Run multi-step agent tasks during a fixed time window without a per-token or per-request charge accumulating.
Custom API Clients: Connect a script, SDK, editor extension, or other client by configuring its API base URL and key.
Image-Based Prompts: Send image-containing requests to a supported model for workflows that need visual input, with up to five images per request.
Open Model Experiments: Try the hosted Qwen3.8-27B model for tasks that use its vision, reasoning, or tool-calling capabilities.

