Skip to content
NataPulse Docs

For Agents

Limited — capability-dependent

For Agents describes how software agents can discover and consume selected NataPulse intelligence.

Product boundary

Section titled “Product boundary”

Agent tools expose only published, read-only product models such as:

  • curated events;
  • clusters;
  • emerging narratives;
  • source-cited reports;
  • other explicitly listed public tools.

They do not expose private workspace content, unpublished events, internal review scores, secrets, raw provider tables, or operational traces.

MCP

Section titled “MCP”

The agent surface uses the Model Context Protocol over a stateless HTTP transport. Machine-readable discovery is available through the published agent documentation files and tool catalogue.

The current configured endpoint is:

POST https://mcp.nata.my/mcp

Clients should retrieve the live tool list rather than rely on a copied static price or schema table.

Access and payment status

Section titled “Access and payment status”

The architecture supports per-call x402 payment using a wallet as the payer identity. Payment activation is operationally gated. Until the live page states that settlement is active, tool access may require approval or another enabled access path.

Discovery and health checks can be available independently from paid tool execution.

Safety and metering

Section titled “Safety and metering”

Agent access can apply:

  • per-call pricing;
  • wallet or identity quotas;
  • request-rate limits;
  • replay protection;
  • read-only tool contracts;
  • publication and field-whitelist gates.
Section titled “Recommended client behavior”

An agent should:

  1. initialize and retrieve the current tool list;
  2. validate the tool schema;
  3. request only the required scope and time window;
  4. preserve citations and timestamps;
  5. distinguish empty results from errors;
  6. disclose that NataPulse output is research information;
  7. avoid converting a signal into an autonomous trade instruction.