diff --git a/README.md b/README.md
index 139b9f1..99a8f79 100644
--- a/README.md
+++ b/README.md
@@ -51,7 +51,7 @@ Phoenix is built for the people actually running agents today — self-hosted, h
 - **Reference-first workflows** — `phoenix://` URIs replace raw values in configs and scripts
 - **Exec credential stripping** — `phoenix exec` injects secrets and strips broker credentials from the child process
 - **Works with existing setups** — `phoenix import` pulls from `.env` files in one command; 1Password users can migrate secrets or broker reads at runtime without moving anything
-- **Agent-native integrations** — built-in MCP server for Claude Code / Claude Desktop, Python/Go/TypeScript SDKs, OpenClaw SecretRef exec provider and plugin-tool path, direct HTTP API
+- **Agent-native integrations** — built-in MCP server for Claude Code / Claude Desktop, Python/Go/TypeScript SDKs, OpenClaw SecretRef exec provider and [openclaw-phoenix](https://github.com/phoenixsec-dev/openclaw-phoenix) runtime plugin, direct HTTP API
 - **Emergency offline access** — break-glass secret retrieval directly from disk when the server is down
 
 Two binaries (`phoenix` client + `phoenix-server`), single codebase, no external runtime dependencies.
diff --git a/docs/integrations.md b/docs/integrations.md
index 75e66f9..39d0c81 100644
--- a/docs/integrations.md
+++ b/docs/integrations.md
@@ -210,7 +210,7 @@ captured and is never written to audit logs.
 
 ### 2. Plugin tools for agent/tool runtime access
 
-Use the separate `openclaw-phoenix` plugin for agent/tool-time workflows:
+Use the separate [`openclaw-phoenix`](https://github.com/phoenixsec-dev/openclaw-phoenix) plugin for agent/tool-time workflows:
 
 - `phoenix_resolve` for Phoenix-aware runtime resolution
 - `phoenix_list` for listing visible paths