Meldport

Choose a model API for your coding agent

The cheapest usable model must speak your agent's protocol and complete the task. Compare token cost after compatibility, tools, and streaming have been checked.

Vendor documentation checked 2026-11-02. These are documentation reviews, not paid end-to-end integration tests.

  • Claude Code

    Direct connection unavailable. Claude Code's gateway contract requires Anthropic Messages. Meldport currently documents Chat Completions, so replacing the base URL is insufficient.

  • Cursor

    Custom endpoint requires validation. Cursor supports provider API keys for chat. A custom endpoint must match the API format Cursor expects; this is not a promise that every Cursor feature uses your gateway.

  • OpenCode

    Custom Chat Completions configuration. OpenCode documents custom OpenAI-compatible providers using the Chat Completions adapter. Configure the provider and model explicitly rather than assuming model discovery.

  • Cline

    OpenAI Compatible configuration guide; validate the client and model combination. Cline provides an OpenAI Compatible provider setting for a custom API base URL, key and model ID. Use that mode for Meldport's Chat Completions contract; the native OpenAI/Codex or Anthropic modes can require different protocols.

  • OpenClaw

    Configuration per vendor documentation; not end-to-end tested. OpenClaw documents custom providers with an explicit openai-completions adapter. The endpoint origin and model capabilities remain part of the integration contract.

  • Codex

    Direct connection unavailable for current documented protocol. The current Codex provider reference supports the Responses wire protocol. Meldport's public contract currently documents Chat Completions instead.

Budget per completed task

A workload with 100,000 uncached input tokens and 10,000 output tokens costs 0.1 × input rate + 0.01 × output rate, with rates quoted per million tokens. This is a calculation, not a benchmark. Include retries and tool-loop turns in your token totals.

Compare published rates and calculate a task budget. The shared free allowance is available on eligible :free aliases; agent compatibility and request limits still need checking before running a tool loop.