Claude Code bug silently replaces MCP tool results with unreadable <<ccr:...>> placeholder
zlotnika · ghdev · 2026-09-26
A newly filed Claude Code issue reports MCP tool results being silently replaced in model context by a placeholder like <<ccr:<hash>,html,NNNB>>, leaving the model unable to read actual output.
Key findings:
- Occurs across two unrelated MCP servers (Supabase and Slack) in one session, ruling out a server-side fault
- Deterministic minimal repro: Supabase querylogs with a trivial two-value query returns the same html,530B marker on repeat calls
- Rules out auth (sibling tool executesql works), payload size (30 bytes vs reported 530B), and permanence (same tool worked earlier in the session); Slack format=full triggers it while format=namesonly renders fine
- The html type marker suggests HTML payloads are stored as references instead of being passed through
The reporter asks that tool output either reach the model or that the model be told explicitly why output was withheld.
More from coding & agent
- Adobe MAX: MCP across Creative Cloud, AI agents editing layers, Firefly Builder and Graph — umesh_ai · 2026-11-11
- Sydney Codex Community Buildathon set for Nov 14 with Cerebras-backed ultrafast compute track — paw_lean · 2026-11-11
- shadcn's improve hits 10k stars: let top models plan, cheap models execute code fixes — shadcn · 2026-11-11
- OpenAI open-sources Codex Security CLI and SDK for AI-driven vulnerability scanning and fixing — openai · 2026-11-11
- Claude Opus 5.5 runs an architecture search and finds AttnRes-Canon, a new top Transformer design — SonglinYang4 · 2026-11-11
- 5 research habits worth knowing: read papers backwards, reproduce first, wire coding agents via MCP — cneuralnetwork · 2026-11-11