diff --git a/AGENTS.md b/AGENTS.md
index 301469e..d7a216c 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -94,7 +94,11 @@ regexes over heavyweight parsers, small focused modules).
 
 - **Do not commit to `main` directly.** Create a feature branch per change and
   open a PR into `main`.
-- Commit identity: `user.email = noreply@anthropic.com`, `user.name = Claude`.
+- **No AI-assistant / Claude Code attribution** anywhere — no `Co-Authored-By`,
+  `Claude-Session`, `🤖 Generated with …`, or `claude.ai/code` in commit
+  messages or PR descriptions, and don't attribute commit authorship to an
+  assistant. See [CLAUDE.md](CLAUDE.md). (The Anthropic *LLM provider* stays —
+  it's a feature.)
 - `.env`, `.venv`, `__pycache__`, `~/.cache/...` and build artefacts are
   gitignored — never commit them.
 - **CI** is one workflow per concern under `.github/workflows/`: `lint.yml`
diff --git a/CLAUDE.md b/CLAUDE.md
new file mode 100644
index 0000000..7a9d65b
--- /dev/null
+++ b/CLAUDE.md
@@ -0,0 +1,19 @@
+# ConfLens — project rules for agents
+
+See [AGENTS.md](AGENTS.md) for contributor/agent guidance and
+[docs/ARCHITECTURE.md](docs/ARCHITECTURE.md) for the design.
+
+## Attribution policy (persistent — applies to every session and agent)
+
+Do **not** add AI-assistant / "Claude Code" attribution anywhere in this
+repository or its GitHub artifacts:
+
+- **No** `Co-Authored-By:` trailers, `Claude-Session:` lines, `🤖 Generated
+  with …` footers, or `claude.ai/code` links in **commit messages** or
+  **pull-request descriptions**.
+- Keep source, docs, and config free of Claude Code / assistant references.
+- Do not attribute commit authorship to an AI assistant.
+
+**Exception:** Anthropic / Claude *as an LLM provider* is a genuine product
+feature (the `anthropic` provider and `claude-*` model ids in `llm.py`) and
+stays — this policy is about tooling attribution, not the app's providers.