Inject request-scoped context (user identity, database handles, session) into ToolLoopAgent tools via experimental_context, so the model never supplies sensitive identity in tool inputs.
Requirements, wiring steps, and what this pattern adds to your project.
Pick how you want to pull this pattern in. Then wire env vars and routes the same way.
Download → Next.js app
Use Download → Next.js app in the preview toolbar. You get a zip scaffold (Pro downloads rewrite preview `/view/.../api/*` paths to `/api/*` automatically).
Extract the zip and install dependencies
Unzip the archive, then run `pnpm install` (or npm / yarn) in the project folder.
Fill .env.local
The download includes .env.local placeholders when needed. Add real values for: OPENAI_API_KEY.
Customize the agent and tool files
Adapt prompts, tools, and stop conditions in the generated project for your product.
Run the app and open the pattern route
Run `pnpm dev`, open the setup/checklist page if present, then navigate to the generated pattern route and verify responses.
OPENAI_API_KEYGet keyAI SDK APIs
ToolLoopAgentexperimental_contexttoolstepCountIsgatewayProviders
app/page.tsxapp/layout.tsxapp/api/context/route.tscomponents/context-demo.tsxcomponents/tool-call-trace.tsxlib/context-agent.tslib/request-context.tslib/demo-users.tstools/get-orders.tstools/get-profile.tsREADME.mdnpm packages
6
Registry components
tools/get-orders.tsCore pattern: tool whose inputSchema omits userId; execute reads identity and db from the second arg's experimental_context, casting to a RequestContext type.
context/route.tsResolves identity server-side and passes it to agent.generate({ prompt, experimental_context }). Returns tool calls and results for inspection.
lib/request-context.tsDefines the RequestContext type (userId, user, db) and a fake database. Replace with your real session + data layer.
Inject request-scoped context (user identity, database handles, session) into ToolLoopAgent tools via experimental_context, so the model never supplies sensitive identity in tool inputs.
Requirements, wiring steps, and what this pattern adds to your project.
Pick how you want to pull this pattern in. Then wire env vars and routes the same way.
Download → Next.js app
Use Download → Next.js app in the preview toolbar. You get a zip scaffold (Pro downloads rewrite preview `/view/.../api/*` paths to `/api/*` automatically).
Extract the zip and install dependencies
Unzip the archive, then run `pnpm install` (or npm / yarn) in the project folder.
Fill .env.local
The download includes .env.local placeholders when needed. Add real values for: OPENAI_API_KEY.
Customize the agent and tool files
Adapt prompts, tools, and stop conditions in the generated project for your product.
Run the app and open the pattern route
Run `pnpm dev`, open the setup/checklist page if present, then navigate to the generated pattern route and verify responses.
OPENAI_API_KEYGet keyAI SDK APIs
ToolLoopAgentexperimental_contexttoolstepCountIsgatewayProviders
app/page.tsxapp/layout.tsxapp/api/context/route.tscomponents/context-demo.tsxcomponents/tool-call-trace.tsxlib/context-agent.tslib/request-context.tslib/demo-users.tstools/get-orders.tstools/get-profile.tsREADME.mdnpm packages
6
Registry components
tools/get-orders.tsCore pattern: tool whose inputSchema omits userId; execute reads identity and db from the second arg's experimental_context, casting to a RequestContext type.
context/route.tsResolves identity server-side and passes it to agent.generate({ prompt, experimental_context }). Returns tool calls and results for inspection.
lib/request-context.tsDefines the RequestContext type (userId, user, db) and a fake database. Replace with your real session + data layer.
Loading preview…
Early-bird pricing — limited time


Open on desktop for the interactive preview.
Inject request-scoped context (user identity, database handles, session) into ToolLoopAgent tools via experimental_context, so the model never supplies sensitive identity in tool inputs.
Loading preview…
Early-bird pricing — limited time


Open on desktop for the interactive preview.
Inject request-scoped context (user identity, database handles, session) into ToolLoopAgent tools via experimental_context, so the model never supplies sensitive identity in tool inputs.