diff --git a/decisions/0007-evidence-chain-postpone.md b/decisions/0007-evidence-chain-postpone.md
new file mode 100644
index 0000000..01bbd8b
--- /dev/null
+++ b/decisions/0007-evidence-chain-postpone.md
@@ -0,0 +1,42 @@
+# ADR-0007 · Evidence chain is separate workstream, not blocking demo
+
+**Status**: accepted
+**Date**: 2026-05
+
+## Context
+
+A "legal-grade evidence chain" — clips with cryptographic manifests, NTS-anchored timestamps, transparency-log replication, eIDAS-compliant TSA signatures — is a strong differentiator and a future moat. But it's also:
+
+- Multi-month engineering effort.
+- Requires legal review per jurisdiction.
+- Has nuanced UX implications (key custody, signature verification flow).
+
+The pressure to "have evidence chain in the demo" risks pulling resources from getting the **basic forensic experience** working end-to-end.
+
+## Decision
+
+Treat evidence chain as a **separate workstream** (Epic 7 — Hardening). For the demo path (Epics 0-6), implement only:
+
+- SHA256 of clips at write time.
+- Manifest JSON with `cam_id`, `sha256`, `ts_local`, `ts_nts`, `model_sha`.
+- Simple signature with a Cell-local key.
+
+That gives a story to tell ("we hash and sign every clip") without the full eIDAS / RFC 3161 TSA ceremony.
+
+The full evidence chain — transparency log, BYOK customer encryption, third-party verifiable manifests, FIDO2-attested operator actions — is **post-MVP**.
+
+## Consequences
+
+**Good**:
+- MVP demo viable end of sprint 4 instead of end of sprint 8.
+- Evidence chain workstream can move at its own (legal-paced) tempo.
+- Customers can adopt MVP first and upgrade to evidence chain later.
+
+**Bad / trade-offs**:
+- Sales conversations with legal-grade buyers (police, banks, insurance) need to clearly say "evidence chain coming in v2" — risk of losing those leads if they need it now.
+- Some early demos may overpromise; need disciplined messaging.
+
+## Alternatives considered
+
+- **Full evidence chain in MVP**: rejected, blocks first revenue by 4-6 months.
+- **No evidence chain ever, position as operational tool only**: rejected, removes a key strategic differentiator.