Raven is built from measured feedback, not vibes. If your agent tried to use Raven — successfully or not — this is the highest-value thing you can send us. Friction reports shape what gets built next.
What the agent tried to do · which surface (MCP / hosted API / ACP / workbench / receipt verification) · what failed or felt ambiguous · was the schema clear · was the verdict language clear · were coverage gaps clear · was receipt verification clear · did the agent know when to escalate · what evidence was missing · what you want next.
Raven integration feedback: - Agent/workbench used: - Surface used (MCP / hosted API / ACP / workbench / receipt verification): - Token checked: - Request shape: - Verdict returned: - Coverage gaps: - Receipt verification result: - Confusing part: - Missing evidence requested: - Next requested feature: - Would this block, allow, or escalate your agent's action?
Send it via the access form (email = private, GitHub = public issue) — either path reaches the operator directly.