opencode drains Claude 5-hour plan limit 1.8x faster via SDK entrypoint label bug
Salloxy · ghdev · 2026-11-02
- A GitHub issue on opencode reveals that using Claude through the Agent SDK consumes the 5-hour Claude plan limit about 1.8x faster than running the same work in Claude Code's own TUI. The cause: the SDK sets CLAUDECODEENTRYPOINT=sdk-ts, and requests labeled sdk-ts/sdk-cli use more of the 5-hour window for identical token usage (weekly limits unaffected).
- Upstream anthropics/claude-code#97074 includes a controlled A/B test showing 1.8x; T3 Code users independently measured 14-15 calls per 1% with sdk-ts vs 25-27 with non-sdk- labels.
- The common settings.json env workaround doesn't work since opencode loads no /.claude settings (settingSources: []). Only exporting CLAUDECODEENTRYPOINT=opencode in the shell before launch helps; the issue proposes fixing it in opencode's ISOLATIONENV.
More from coding & agent
- From chatbot to agent: how tools, planning and memory were added step by step — _jaydeepkarale · 2026-11-04
- Binance's $7B hack reportedly started with a tampered Homebrew install during vibe coding — xuanalogue · 2026-11-04
- $100 of GPU time edits model weights to cut thinking tokens by 40% — jarrodwatts · 2026-11-04
- Cua Driver integrates with Google Antigravity for background cross-app computer agents — m4rkmc · 2026-11-04
- Sentdex: OpenHands was the first reasonably useful agent to abstract away the code editor — rajistics · 2026-11-04
- Security advisory: run AI agents in managed cloud environments, not employee endpoints — amasad · 2026-11-04