AI Agents on z4p

z4p does not embed a single LLM. It is a harness: it composes context from the vault bible, runs tools and loops, proposes actions, and waits for approval.

The actual intelligence can come from any of the 50+ providers z4p supports (OpenAI, Anthropic, local models, etc.).

How agents work

1. Compose — Vault bible + live memory + learnings
2. Retrieve — FTS5 search for relevant past outcomes
3. Decide — LLM chooses tools and proposes actions
4. Simulate — Routes are dry-run against venues
5. Approve — Trader approves before execution

Planned in z4p Mobile

The companion app will surface specialized agents for:

• Portfolio analysis
• Market research
• Trade rationale explanations
• Opportunity scanning

These agents prepare actions; the harness executes only after approval.