@orangecheck/agent-vercel

Wrap Vercel AI SDK tool() invocations with OC Agent scope enforcement and emit a signed agent-action envelope per tool execution. Provider-agnostic — works under Anthropic, OpenAI, Cohere, or any other model the AI SDK speaks.

Why

The Vercel AI SDK has converged on a clean tool({description, parameters, execute}) primitive that serializes naturally across providers. A single adapter at this layer covers a large fraction of teams shipping production agents.

This module wraps execute so every successful call produces a signed agent-action envelope citing the active delegation:

• Pre-call: scope check (vercel:tool(verb=<verb>) against the delegation; refuses if not admissible).
• Canonicalize the (verb, args, callId) tuple, BIP-322 sign as the agent address, emit envelope.
• Run the real handler.
• Return both the result (back to the model via the SDK) and the envelope (for your audit pipeline).

Status

v0.0.1 · in design. API shape is stable; the canonicalization layout ({args, call_id, verb} lexicographic, RFC 8785 canonical JSON, trailing LF) is locked. Production wiring with specific AI-SDK releases is intentionally not baked in — ocTool() is provider/SDK-agnostic so you can call it from any release of ai you're already on.

Install

npm i @orangecheck/agent-vercel
# peer deps:
npm i @orangecheck/agent-core @orangecheck/agent-signer

Quickstart

import { tool } from 'ai';
import { ocTool } from '@orangecheck/agent-vercel';

const invoiceCreate = ocTool({
    verb:        'invoice.create',
    description: 'create a new invoice',
    parameters:  invoiceSchema,
    execute:     async (args) => myInvoiceCreateImpl(args),
});

const tools = {
    'invoice.create': tool({
        description: invoiceCreate.description,
        parameters:  invoiceCreate.parameters as never,
        execute: async (args, { toolCallId }) => {
            const { result, action } = await invoiceCreate.execute(args, {
                agent, delegation, callId: toolCallId,
            });
            await yourAuditPipeline.append(action);
            return result;
        },
    }),
};

const { text } = await generateText({
    model: anthropic('claude-sonnet-4-6'),
    tools,
    prompt: 'invoice acme for $14.20',
});

The pattern: ocTool defines the OC-Agent-shaped wrapper; the AI SDK's tool() calls into that wrapper plus your audit pipeline.

What gets stamped

Stamp's content.hash is a SHA-256 of:

{
  "args":    <canonical JSON of args>,
  "call_id": "<sdk-assigned tool call id>",
  "verb":    "<verb>"
}

Default scope_exercised is vercel:tool(verb=<verb>). Override via scopeExercised.

API

• canonicalizeToolCall(call) — RFC 8785 canonical bytes.
• toolCallHash(call) — sha256:<64-hex>.
• stampToolCall(input) — sign without executing.
• ocTool({verb, parameters, description, execute}) — provider-agnostic wrapped tool.

License

MIT.

Related

• oc-agent-protocol — normative wire format.
• @orangecheck/agent-core — verifier library.
• @orangecheck/agent-mcp / -anthropic / -openai — sibling adapters.
• fleet.ochk.io/integrations/vercel-ai — managed-tier story.