# Uplink (npm · @uplink-code/mcp)

Automate a real, logged-in browser on your own device — no stored credentials, no bot detection.

- Trust score: 67/100 (medium)
- Change this week: +16
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-08-03

## Components

- npm · `@uplink-code/mcp`: 67/100 (this document), [markdown](https://verifymcp.io/servers/build-uplink-uplink/uplink-code-mcp.md), [page](https://verifymcp.io/servers/build-uplink-uplink/uplink-code-mcp)

## Channel facts

- Registry: `npm`
- Package: `@uplink-code/mcp`
- Version: `0.1.4`
- Transport: `stdio`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-08-03.

- **Supply Chain Security**: 88/100
  - No malware found by supply-chain analysis.
  - Only part of the dependency tree could be resolved (105 of 109), so this covers what we could see, not the whole tree.
  - No install/post-install scripts declared.
  - Only part of the dependency tree could be resolved (105 of 109), so this covers what we could see, not the whole tree.
- **Provenance & Transparency**: 32/100
  - Source repository is publicly reachable at the declared URL.
  - Provenance check failed: no build-provenance attestation is published.
  - License check failed: the license (proprietary) isn't a recognized OSI-approved license.
  - Actively maintained (last published 40 days ago).
  - Disclosure check failed: no security disclosure policy was found in the source repository.
- **Schema Quality & AI Usability**: 78/100
  - 100% of prompts and resources have a non-trivial description (not blank, and not just the item's name).
  - AI-judged instruction clarity (excellent).
  - Context-footprint check failed: tool/resource definitions use about 3367 tokens (~187/item across 18 items; 15 tools + 3 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 27/100
  - Stability observed for 8 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 100/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 100% of tool parameters carry a description.
- **Capabilities**: 100/100
  - Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.

## Install

### Claude

```bash
claude mcp add build-uplink-uplink -- npx -y @uplink-code/mcp
```

### Codex

```bash
codex mcp add build-uplink-uplink -- npx -y @uplink-code/mcp
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "build-uplink-uplink": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "@uplink-code/mcp"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add build-uplink-uplink --command npx --arg -y --arg @uplink-code/mcp
```

### Hermes

```yaml
mcp_servers:
  build-uplink-uplink:
    command: "npx"
    args: ["-y", "@uplink-code/mcp"]
```

### Other

```json
{
  "mcpServers": {
    "build-uplink-uplink": {
      "command": "npx",
      "args": [
        "-y",
        "@uplink-code/mcp"
      ]
    }
  }
}
```

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-08-02 (score 67, +31)

- [security regression] Provenance: unverified → fail
- [security improvement] Install scripts: unverified → pass
- [security improvement] Known CVEs: unverified → partial
- [security improvement] Malware scan: unverified → pass
- [security] Stability: Stability not yet verified: we do not have a sandbox capture of the MCP schema this version of the package serves yet.
- [functional regression] License: unverified → fail
- [functional regression] Schema quality: 100 → unverified
- [functional regression] Security disclosure: fail → unverified
- [functional regression] Capabilities: pass → unverified
- [functional regression] Tool coverage: 100 → unverified
- [functional improvement] Stability: unverified → 0.23
- [functional improvement] Maintenance: unverified → pass
- [functional improvement] Dependency health: unverified → partial
- [functional] Licence: proprietary

### 2026-08-01 (score 36, −5)

- [functional] We updated how we score, so this day's move reflects our rubric, not a change to the server

### 2026-07-31 (score 41, −10)

- [security regression] Malware scan: pass → unverified
- [functional improvement] Schema quality: unverified → excellent

### 2026-07-27 (score 51)

First indexed and scored.

## MCP tools (15)

### `uplink_connect` (~421 tokens)

Start a new Uplink session for pairing a device via the Uplink Connect app. Call this before any other tool that drives the device.

AFTER this returns, in the SAME response turn, IMMEDIATELY call uplink_wait_for_device — do NOT end your turn saying "let me know when you've paired" or wait for the user to confirm. wait_for_device is the pairing-detection mechanism; see its description for the timeout-and-refresh loop. Never poll uplink_session_status as a substitute.

CRITICAL: Claude Code collapses ALL MCP tool output behind ctrl+o — the tool result is invisible unless expanded, so your response prose is the only thing the user sees. Render the pairing instructions in YOUR OWN prose, using the exact link and code from the tool result.

Pairing hands off from the user's desktop to their phone, so LEAD WITH THE LINK: on iPhone, opening the link (or scanning a QR) launches the Connect App Clip and pairs instantly, no install. Give the 6-digit code as the SECONDARY path (Android, or when Connect is already open). Render roughly:
   \```
   ^ Uplink session started! Pair your phone:
   • iPhone — open this link on your phone to pair instantly (Connect App Clip, no install): http://localhost:4100/link/s/<uuid>
   • Android / Connect already open — enter code 123-456 (expires in 30s)
   \```
   If the tool result includes a "Link copied to your clipboard" line, surface it too — the user can paste the link onto their iPhone via Universal Clipboard.

The user can also scan a QR — mention they can ask for one (only call uplink_get_qr if they explicitly ask).

If the user mentions a custom SDK implementation, also include the JWT line from the tool result.

Do not paraphrase or omit the link/code — print them verbatim.

Input parameters:

- `projectId` (string): Uplink project ID. Defaults to the API key owner's Playground project.

### `uplink_refresh_pairing_code` (~111 tokens)

Mint a fresh 6-digit pairing code for the active session. Use this if the previous code expired (30s TTL) before the user typed it — the session itself stays valid, only the OTP rotates. Surface the new code in YOUR OWN prose, same format as uplink_connect (XXX-XXX), e.g.:
   \`New pairing code: 123-456 (expires in 30s)`
Claude Code collapses tool output; the code must be visible in your prose.

### `uplink_get_qr` (~159 tokens)

Render the active session's pairing URL as an ASCII QR code. Opt-in: only call when the user explicitly asks to see a QR (e.g., 'show me a QR' / 'I want to scan').

CRITICAL: Claude Code collapses tool output behind ctrl+o. After calling this tool, you MUST paste the returned QR ASCII art VERBATIM into your own response prose — every line of unicode half-blocks, exactly as returned. Do not say 'scan above' or 'see the QR' — those refer to the collapsed tool output which the user cannot see. The user will only see what you write directly. Yes, this is verbose, but the user explicitly asked for the QR; render it.

### `uplink_wait_for_device` (~273 tokens)

Block until a device finishes pairing on the active session, then return its metadata (deviceModel, platform, platformVersion, deviceType). Resolves immediately if a device is already paired.

Default timeout is intentionally short (25s, under the 30s OTP TTL) so each call returns control to you in time to refresh the code before it expires. When this errors with `pairing_timeout` and the user hasn't indicated they're done:

1\. Call `uplink_refresh_pairing_code` to mint a fresh code.
2\. Surface the new code in YOUR OWN prose (same XXX-XXX format as uplink_connect):
   \`New pairing code: 123-456 (expires in 30s)`
3\. Call `uplink_wait_for_device` again to keep waiting.

Keep looping until either (a) the call returns successfully (device paired), (b) the user says to stop, or (c) the user says they've paired (still call wait_for_device once more — it returns immediately if the device is already there). This loop gives the user windows between iterations to interrupt or change approach.

Input parameters:

- `timeout_ms` (number): How long to wait for a device to pair. Default 25000 (25s, intentionally under OTP TTL). Max 600000 (10 min).

### `uplink_session_status` (~25 tokens)

Return whether a session is active and whether a device is paired to it.

### `uplink_get_current_url` (~22 tokens)

Returns the current URL of the device's active page.

### `uplink_get_dom_snapshot` (~45 tokens)

Snapshot the DOM of the active device. Returns the body tree (formatted for LLM consumption) plus head metadata (title, base, meta tags, link rels).

### `uplink_evaluate_javascript` (~68 tokens)

Run a JS expression in the device's page context and return the result. Can take either a single expression (`document.title`) or a statement block (`const r = await fetch(...); return r.json()`).

Input parameters:

- `expression` (string, required): JS expression or statement block to evaluate

### `uplink_await_network_request` (~81 tokens)

Wait for the next network request whose URL contains `url_pattern` (substring match). Returns method, URL, status, headers, and a body preview.

Input parameters:

- `timeout_ms` (number): How long to wait. Default 15000, max 60000.
- `url_pattern` (string, required): Substring to match against the request URL

### `uplink_run_script` (~90 tokens)

Execute a script against the device. The script body has `page` and `console` bound. Returns counts of captured flows + logs plus any error.

Input parameters:

- `code` (string, required): JS source to execute. `page` and `console` are bound in scope.
- `wait_for_idle_ms` (number): Settling window after script resolves; default 3000, max 10000.

### `uplink_read_logs` (~65 tokens)

Return captured console logs from `uplink_run_script` invocations on the active session. Optional `level` filter and `limit`.

Input parameters:

- `level` (string): Filter by log level.
- `limit` (number): Most-recent N entries. Defaults to all retained.

### `uplink_search_network` (~75 tokens)

Search captured network requests by URL substring. Returns matches with method, URL, status, headers, and body previews. Use uplink_get_network_request to fetch a single entry by id.

Input parameters:

- `limit` (number): Most-recent N matches.
- `url_pattern` (string, required): Substring to match against the request URL.

### `uplink_get_network_request` (~38 tokens)

Fetch a single captured network request by id (the id from uplink_search_network).

Input parameters:

- `id` (number, required): Network request id.

### `uplink_list_commands` (~74 tokens)

List captured page commands (method, args, result/error, timing) emitted while uplink_run_script executed. Optional `method` filter and `limit`.

Input parameters:

- `limit` (number): Most-recent N entries.
- `method` (string): Filter by page method name (e.g., "goto", "click").

### `uplink_get_command_result` (~37 tokens)

Fetch a single captured page command by id (the id from uplink_list_commands).

Input parameters:

- `id` (number, required): Command id.

## Diagnostics

Captured diagnostic sections: Provenance, Dependencies. The full working is on the page: https://verifymcp.io/servers/build-uplink-uplink/uplink-code-mcp#diagnostics

## Score history

- 2026-08-03: 67
- 2026-08-02: 67
- 2026-08-01: 36
- 2026-07-31: 41
- 2026-07-29: 51
- 2026-07-28: 51
- 2026-07-27: 51

## Links

- npm package: https://www.npmjs.com/package/@uplink-code/mcp
- Socket report: https://socket.dev/npm/package/@uplink-code/mcp
- Repository: https://github.com/uplink-code/mcp
- Website: https://docs.uplink.build/
- Changelog RSS feed: https://verifymcp.io/servers/build-uplink-uplink/uplink-code-mcp/changelog.xml
- Changelog JSON feed: https://verifymcp.io/servers/build-uplink-uplink/uplink-code-mcp/changelog.json
- HTML version of this page: https://verifymcp.io/servers/build-uplink-uplink/uplink-code-mcp