Bring your own capacity
OpenManifest never holds upstream tokens. Your Copilot subscription supplies the model capacity; we forward requests and observe usage.
Documentation
Get any agent running in two minutes.
Pick the harness you use, drop in your omk_live_* key, and you're done.
Quickstarts
Claude Code
Open →Anthropic Messages API. Drop in ANTHROPIC_BASE_URL and go.
ANTHROPIC_BASE_URL="https://api.openmanifest.one/v1" \
ANTHROPIC_API_KEY="omk_live_…" \
claudeCodex
Open →OpenAI Chat Completions or Responses API.
OPENAI_BASE_URL="https://api.openmanifest.one/v1" \
OPENAI_API_KEY="omk_live_…" \
codexOpenCode
Open →OpenAI-compatible. Same env vars as Codex.
OPENAI_BASE_URL="https://api.openmanifest.one/v1" \
OPENAI_API_KEY="omk_live_…" \
opencodeCursor
Open →Settings → Models → OpenAI Base URL.
Settings → Models →
OpenAI Base URL: https://api.openmanifest.one/v1
API key: omk_live_…Concepts
Three protocols, one upstream
We accept OpenAI Chat Completions, Anthropic Messages, and the OpenAI Responses API. All three are translated to Copilot's format and back, with real chunk-incremental streaming.
Edge-deployed
The gateway runs at the network edge. Requests stay there; only the upstream call to api.githubcopilot.com leaves it.