API integrations and tools

Use this page when the main problem is API wiring, tool calling, browsing integration, SDK usage, or external system integration.

Use this page when the issue is how the model connects to tools or other systems: API calls, tool calling, web browsing, SDK setup, or integration review.

Prerequisites

Before you start

• Know whether the task is really an integration problem, not just prompt design or assistant-product behavior.
• Know which API, SDK, or external tool surface is involved.
• Decide whether the task really needs live public information before adding web access.

Scope

When to use API integrations and tools

• Use this page when the main problem is connecting model output to APIs, tools, web access, or other external systems.
• Use it when you need to check an integration against the provider documentation.
• Use it when you need to choose the right review path before changing an integration.
• Use it when the work is about connections and external behavior, not just prompt design inside one interface.

Use another topic when the main problem is elsewhere:

• Prompt engineering — The main problem is how to write the prompt, not how the model connects to other systems.
• AI assistants — The main problem is how a chat assistant product behaves.
• Agents and workflows — The main problem is routing, orchestration, memory across steps, or multi-step system design.

Continue with the right resource

Choose the resource that matches the next action, not just the content type.

Guides

Use this when the integration path is already chosen and the next step is a structured engineering review.

• Run the engineering quality gate for an integration change Use this when the next step is a structured engineering review of an integration change, not initial scoping.

Policies

Use these when integrations must follow fixed rules about provider docs, sources, and engineering review.

• Require an implementation review against the vendor docs Use this when integration decisions must be checked against provider or standards documentation.
• Require the engineering quality gate for integration changes Use this when an integration must pass a structured engineering review before changes are accepted.
• Require web checks for current or niche claims Use this when the integration depends on recent or specialized public information.
• Require the right review path for an integration change Use this when the team must choose the correct review mode before changing code, architecture, or integration behavior.

Prompts

Use these reusable prompts when you want to run integration review, web checks, or citation control through the prompt layer.

• System prompt: check the implementation against the vendor docs Use this to enforce provider-doc checks during implementation review.
• Workflow prompt: run a vendor-doc implementation review Use this when you want a guided implementation review tied to provider documentation.
• System prompt: require web checks and citations Use this when the answer depends on current or specialized public information.
• Workflow prompt: request web browsing Use this when you want a controlled browsing request inside an integration workflow.
• Workflow prompt: return citations in a fixed format Use this when the final answer must present citations in a fixed structure.

Reference

Use these when you want the background behind prompt-layer boundaries, web-tool risk, and integration review decisions.

• Reference: how prompt layers and policy layers affect integrations Use this to understand how instruction layers and policy layers interact in integration-facing systems.
• Reference: where web-tool prompt-injection risk appears Use this to reason about prompt-injection exposure when web tools are attached to an LLM workflow.
• Reference: CODE-RPC for GPT (OpenAI/ChatGPT) Use this when you need the operating model behind structured technical review for GPT-based integration work.

Articles

Use these when you need deeper analysis of control-plane failures, prompt-assembly enforcement, and trust boundaries at the integration layer.

• Read Control-Plane Failure Patterns in Tool-Using LLM Systems Long-form analysis of control-plane failure modes in tool-using LLM systems.
• Read Prompt Assembly Policy Enforcement: Typed Provenance to Prevent Authority Confusion Long-form analysis of provenance control and prompt-assembly enforcement at the integration layer.
• Read The attack surface starts before agents: the LLM integration trust boundary Long-form analysis of the first trust boundary at the LLM integration layer.