# How do I verify Vercel AI SDK actions with Provely?

> Connect Vercel AI SDK to Provely through the MCP server, the CLI, the SDK, the REST API. The guidance files, the hooks, and the CI gate.

Canonical: https://provely.sh/agents/vercel-ai  
Last reviewed: 2026-09-05  

**Vercel AI SDK connects to Provely through the MCP server, the CLI, the SDK, the REST API. The agent does the action. Provely verifies the outcome and signs a receipt. Vercel AI SDK never holds the verifier credentials.**

A person verified the Vercel AI SDK capabilities on 2026-10-06. Agent products change monthly. Re-verify before you build.

Source: https://provely.sh/agents (retrieved 2026-10-06)

## Which surfaces does Vercel AI SDK use?

| Surface | Support | Detail |
| --- | --- | --- |
| [MCP server](/docs/mcp) | Supported | transport stdio, http |
| [CLI](/docs/cli) | Supported | Runs through the shell tool of the agent. |
| [SDK](/docs/sdk) | Supported | languages typescript |
| [REST API](/docs/api) | Supported | Any HTTP call reaches the control plane. |

## Which guidance files does Vercel AI SDK read?

Provely ships one Agent Skill for this host, so the agent knows to verify an outcome before it claims done. The renderer writes prompt fragment from one source. A new agent needs a profile, not core code.

<details>
<summary>Explain: each file, its path and its write mode</summary>

| Format | Path | Mode |
| --- | --- | --- |
| prompt fragment | `provely/prompt-fragment.md` | fragment |
| CI gate workflow | `.provely/ci/github-actions.yml` | fragment |

</details>

## How is a false completion claim blocked?

Guidance asks. A hook and the CI gate enforce. A host with neither still receives the verdict, and a person reads it before the work ships.

<details>
<summary>Explain: what each control does on this host</summary>

| Enforcement | Support | Detail |
| --- | --- | --- |
| Advisory guidance | Supported | The instruction files state the rules for every agent. |
| Lifecycle hooks | Supported | `tool_execute` |
| CI gate | Supported | `provely verify --wait` fails the pipeline unless the verdict is VERIFIED. |

</details>

## How do I connect Vercel AI SDK?

### Connect Vercel AI SDK to Provely

1. **Install the CLI.** Run `npx provely --version`, or download the static binary.
2. **Register the MCP server.** Add the Provely server. Use `provely mcp` for stdio, or the hosted HTTP server with a bearer token. The tools are `begin`, `verify`, `status`, and `receipt`.
3. **Write the guidance files.** Run `provely init --agent vercel-ai`. It writes `provely/prompt-fragment.md` (fragment), `.provely/ci/github-actions.yml` (fragment). The guidance carries one rule: a successful tool call is not completion.
4. **Install the hooks.** Enable the `tool_execute` hooks. A hook calls the CLI and blocks a completion claim while an operation is PENDING.
5. **Add the CI gate.** Run `provely verify --wait <duration>` in the pipeline. The step fails unless the verdict is VERIFIED. Exit codes: 0 VERIFIED, 2 PENDING, 3 FAILED, 4 CONTRADICTED, 5 UNVERIFIABLE, 1 error.
6. **Report the verdict exactly as returned.** VERIFIED: "The action is verified complete. Receipt: <id>." PENDING: "The action is accepted but not yet verified. Operation: <id>."

## What else does the profile state?

- Verified 2026-10-06: this repository ships the interception point. import '@provely/instrumentations/register' guards generateText, streamText, and ToolLoopAgent when a file imports them. The guard checks each tool call before execute runs. Source: examples/typescript/vercel-ai/app.ts.
- The interception point is the register import, not a script. It blocks only when the process loads it before the ai package. A process that never loads it gets advisory guidance and the CI gate.

> Vercel AI SDK keeps its action credentials, and no surface returns the verifier credentials to it. Read [the credential boundary](/docs/security). An operation outlives the session that opened it. Read [the API reference](/docs/api).

### Which host versions does the profile cover?

`>=7.0.0`. The integrations team last verified the profile with ai 7.0.128 and @provely/instrumentations 0.1.0 on 2026-10-06.

## Read next

- [Verify Claude Code actions](https://provely.sh/agents/claude-code)
- [Verify Codex actions](https://provely.sh/agents/codex)
- [Verify Anthropic Claude API actions](https://provely.sh/verify/anthropic)
- [See every agent integration](https://provely.sh/agents)
- [Read the MCP server reference](https://provely.sh/docs/mcp)
- [Read the credential boundary](https://provely.sh/docs/security)