Skip to content

Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors

Rust

docs(llm): note the Cursor provider's schema-in-prompt and inherited environment

AkitaOnRails00ab34b0d786
diff --git a/docs/llm-providers.md b/docs/llm-providers.md
index 923b3b73..6824bc96 100644
--- a/docs/llm-providers.md
+++ b/docs/llm-providers.md
@@ -82,7 +82,10 @@ ai-memory llm-test --provider codex --model gpt-5.6-luna --structured --prompt "
`AI_MEMORY_CURSOR_AGENT` overrides the binary. The provider writes the prompt
into a temporary workspace and calls `agent --print --mode ask --trust`. It
does not pass `--yolo` or `--force`: consolidation text is untrusted captured
-content, and those flags would allow tool execution.
+content, and those flags would allow tool execution. The Cursor CLI has no
+JSON-schema mode, so the schema travels in the prompt and the answer is parsed
+and validated against it. The `agent` process inherits the server's
+environment, so run the server with only the credentials it needs.
```bash
export AI_MEMORY_LLM_PROVIDER=cursor