# offload (npm · offload-mcp-server)

Hand heavy read-only work to a background agent CLI (Antigravity) and read back a short report.

- Trust score: 74/100 (medium)
- Change this week: +2
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-21

## Components

- npm · `offload-mcp-server`: 74/100 (this document), [markdown](https://verifymcp.io/servers/tailucanh-offload-mcp-server/offload-mcp-server.md), [page](https://verifymcp.io/servers/tailucanh-offload-mcp-server/offload-mcp-server)

## Channel facts

- Registry: `npm`
- Package: `offload-mcp-server`
- Version: `0.4.2`
- 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 2026-11-21.

- **Supply Chain Security**: 97/100
  - No malware found by supply-chain analysis.
  - No known CVEs affecting this package version or its production dependencies.
  - No install/post-install scripts declared.
  - 4 of 9 dependencies flagged as unhealthy.
- **Provenance & Transparency**: 48/100
  - Source repository is publicly reachable at the declared URL.
  - Provenance check failed: no build-provenance attestation is published.
  - Clear OSI-approved license (MIT).
  - Actively maintained (last published 10 days ago).
  - Publishes a security disclosure policy (SECURITY.md).
- **Schema Quality & AI Usability**: 64/100
  - AI-judged instruction clarity (excellent).
  - Context-footprint check failed: tool/resource definitions use about 1393 tokens (~278/item across 5 items; 5 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 43/100
  - Stability observed for 13 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.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 5 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 5 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a current MCP spec version (2026-07-28).

## Install

### How do I install the offload MCP server?

offload runs locally as an npm package, launched with npx -y offload-mcp-server. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add tailucanh-offload-mcp-server -- npx -y offload-mcp-server
```

### Cursor

```json
{
  "mcpServers": {
    "tailucanh-offload-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "offload-mcp-server"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "tailucanh-offload-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "offload-mcp-server"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add tailucanh-offload-mcp-server -- npx -y offload-mcp-server
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "tailucanh-offload-mcp-server": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "offload-mcp-server"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add tailucanh-offload-mcp-server --command npx --arg -y --arg offload-mcp-server
```

### Hermes

```yaml
mcp_servers:
  tailucanh-offload-mcp-server:
    command: "npx"
    args: ["-y", "offload-mcp-server"]
```

### Netclaw

```json
{
  "McpServers": {
    "tailucanh-offload-mcp-server": {
      "Transport": "stdio",
      "Command": "npx",
      "Arguments": [
        "-y",
        "offload-mcp-server"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add tailucanh-offload-mcp-server -t stdio -c npx -a -y offload-mcp-server
```

### Other

```json
{
  "mcpServers": {
    "tailucanh-offload-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "offload-mcp-server"
      ]
    }
  }
}
```

## 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.

### 2026-11-20 (score 74, +1)

No change was recorded against any check on this day. Stability & Change Management went from 37 to 40. That category is still filling its 30-day observation window: 11 days of observed history at the previous scan, 12 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-19 (score 73, 0)

- [functional regression] Dependency health: 0.88 → 0.79

### 2026-11-16 (score 73, +1)

No change was recorded against any check on this day. Stability & Change Management went from 23 to 27. That category is still filling its 30-day observation window: 7 days of observed history at the previous scan, 8 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-14 (score 72, +1)

No change was recorded against any check on this day. Stability & Change Management went from 17 to 20. That category is still filling its 30-day observation window: 5 days of observed history at the previous scan, 6 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-12 (score 71, +1)

No change was recorded against any check on this day. Stability & Change Management went from 10 to 13. That category is still filling its 30-day observation window: 3 days of observed history at the previous scan, 4 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-11 (score 70, +15)

- [security improvement] Malware scan: unverified → pass

### 2026-11-10 (score 55, −14)

- [security regression] Malware scan: pass → unverified
- [functional] Package version: 0.4.1 → 0.4.2

### 2026-11-09 (score 69, +15)

- [security regression] Tool safety: pass → unverified
- [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] Capabilities: pass → unverified
- [functional regression] Tool coverage: 100 → unverified
- [functional improvement] Stability: unverified → 0.03
- [functional] First check of Schema quality: unverified
- [functional] Package version: 0.4.0 → 0.4.1

## MCP tools (5)

### `offload_submit_task` (~921 tokens)

Offload a heavy read-only task

Hand a heavy READ-ONLY task to a background worker agent (default: agy) and get a job_id back immediately.

Use it when the work means reading a lot but the answer is small: scanning many files, filtering large logs, analyzing a big bundle or resource file, inventorying APIs/strings/config across a tree. The worker replies with a short Summary plus Findings with path:line evidence, so you read a few hundred words instead of the files.

Do NOT use it for edits, for small lookups you can do in one or two reads, or for anything security-sensitive you must verify line by line yourself. The worker is another model: treat its findings as leads and spot-check the ones you act on.

Write a complete BRIEF. The worker executes it literally and does nothing else, so anything you leave vague it will skip or guess:
\- task: the goal as one concrete question.
\- steps: ordered, concrete actions; each names what to open or search and what to record.
\- output_hint: the exact shape of the Findings (columns, grouping, sort order, what goes in each column).
\- done_when: the stop condition, so the worker neither stops early nor wanders.
\- context (optional): what you already know (conventions, file roles), so the worker does not rediscover it.
\- exclude (optional): files, folders or kinds of results to leave out.
The worker returns a Step log; the status tells you which steps it did not finish and which in-scope files it never opened.

Each job costs the worker a fixed preamble (about 20k tokens for agy) before it reads anything, so do small lookups yourself. A brief identical to a finished job, over unchanged files, returns that job instead of running again (reuse=false to force a new run).

Next: call offload_get_status with the job_id and wait_sec=50 until it finishes. To ask the worker more about its report, use offload_followup: it keeps what the worker already read.

Example:
task="Which HTTP endpoints does api/v1 register, and what auth does each require?"
steps=["List every .php…

Input parameters:

- `backend` (string): Backend name from the config. Available: agy.
- `context` (string): Facts you already know that save the worker time: conventions, which file does what, names to look for.
- `done_when` (string, required): Stop condition, e.g. "every file under api/v1 is read and every route is listed".
- `exclude` (array): What to leave out: files, folders, or kinds of results, e.g. ["tests/", "commented-out code"].
- `model` (string): Exact worker model id; overrides tier.
- `output_hint` (string, required): Exact shape of the Findings, e.g. "table: name | path:line | purpose, sorted by path".
- `paths` (array): Files or folders to focus on, relative to the workspace, e.g. ["api/core", "dist/app.js"]. Empty = whole workspace.
- `reuse` (boolean): Return a finished job with the same brief over unchanged files instead of running again.
- `steps` (array, required): Ordered actions for the worker, each concrete: what to open or search, what to record. The worker performs exactly these, in order, and logs each as done/partial/not done.
- `task` (string, required): The goal: one concrete question the report must answer, e.g. "Which endpoints in api/v1 skip the CSRF check?"
- `tier` (string): Worker model tier, when the backend defines tiers: low for mechanical inventories, medium by default, high for hard reading or a retry after a weak result.
- `timeout_sec` (integer): Kill the worker after this many seconds. Defaults to the backend's timeoutSec.
- `workspace` (string): Absolute workspace path. Defaults to the current project; other paths must be listed in allowedRoots.

### `offload_followup` (~158 tokens)

Ask the worker more about a finished job

Continue a finished job's worker conversation: ask about its report, or answer the question it ended with (an "ASK:" line). The worker keeps what it already read, so this is much cheaper than a new job. Returns a new job_id; collect it with offload_get_status. Works only with backends that can resume a conversation (agy).

Input parameters:

- `job_id` (string, required): The finished job to continue.
- `message` (string, required): Your question or answer, e.g. "Does api/v1/user/profile.php:26 reject an expired token? Quote the check."
- `output_hint` (string): Shape of the Findings for this answer.
- `timeout_sec` (integer): Kill the worker after this many seconds.

### `offload_get_status` (~124 tokens)

Get offloaded job status

Get the status of one offloaded job, optionally waiting up to 50s for it to finish. When the job is done this already includes the worker's Summary, so you usually do not need offload_get_result.
Without job_id, lists the most recent jobs.

Input parameters:

- `job_id` (string): Job ID from offload_submit_task. Omit to list recent jobs.
- `limit` (integer): How many recent jobs to list when job_id is omitted.
- `wait_sec` (integer): Seconds to wait for completion before answering (max 50).

### `offload_get_result` (~143 tokens)

Read an offloaded job's report

Read the worker's report for a finished job, in pages of up to 25000 characters.
part="summary" (short answer), "findings" (evidence list), "steps" (Step log: done/partial/not done per step), "uncertain", "full" (whole report), "stderr" (backend error output, for debugging a failed job).

Input parameters:

- `job_id` (string, required): Job ID from offload_submit_task.
- `max_chars` (integer): Maximum characters to return.
- `offset` (integer): Character offset, for paging through long parts.
- `part` (string): Which part of the report to read.

### `offload_cancel_task` (~47 tokens)

Cancel an offloaded job

Stop a queued or running offloaded job and its worker process tree. Finished jobs are left as they are.

Input parameters:

- `job_id` (string, required): Job ID from offload_submit_task.

## Diagnostics

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

## Score history

- 2026-11-21: 74
- 2026-11-20: 74
- 2026-11-19: 73
- 2026-11-18: 73
- 2026-11-17: 73
- 2026-11-16: 73
- 2026-11-15: 72
- 2026-11-14: 72
- 2026-11-13: 71
- 2026-11-12: 71
- 2026-11-11: 70
- 2026-11-10: 55
- 2026-11-09: 69
- 2026-11-08: 54

## Common questions

### What is the offload MCP server?

offload is an MCP server listed in the public MCP registry as io.github.tailucanh/offload-mcp-server. Hand heavy read-only work to a background agent CLI (Antigravity) and read back a short report. This page covers its npm package (offload-mcp-server).

### Is the offload MCP server safe to use?

offload scores 74 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 21 November 2026. It declares no install or post-install scripts. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the offload MCP server expose?

offload exposes 5 tools: offload_submit_task, offload_followup, offload_get_status, offload_get_result, offload_cancel_task. Their descriptions and schemas cost roughly 1,393 tokens of context every time the server is loaded.

### Is the offload MCP server still maintained?

offload is still listed as active in the MCP registry. We last reached this channel on 21 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

### What licence is the offload MCP server under?

offload declares the MIT licence, which is OSI-approved. That covers the source only, and says nothing about the cost of any service it calls.

### What did VerifyMCP test on the offload MCP server?

VerifyMCP ran 16 checks on offload, covering Malware scan, Known CVEs, Install scripts, Dependency health and 12 more. Endpoint Security and Transport & Reachability do not apply to an npm package, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.

## Links

- npm package: https://www.npmjs.com/package/offload-mcp-server
- Socket report: https://socket.dev/npm/package/offload-mcp-server
- Repository: https://github.com/tailucanh/offload-mcp-server
- Changelog RSS feed: https://verifymcp.io/servers/tailucanh-offload-mcp-server/offload-mcp-server.xml
- Changelog JSON feed: https://verifymcp.io/servers/tailucanh-offload-mcp-server/offload-mcp-server/changelog.json
- Trust report (JSON): https://verifymcp.io/servers/tailucanh-offload-mcp-server/offload-mcp-server.json
- HTML version of this page: https://verifymcp.io/servers/tailucanh-offload-mcp-server/offload-mcp-server
