grievous-mcp

"Your lightsabers will make a fine addition to my collection."

A hallucination engine. Typed pseudorandom data via Anthropic API. MCP server.

---

⚠️ No Guarantees

This is not a PRNG. This is not a test data framework with schema validation. This is a footgun with good trigger discipline expected of the operator.

• Output is plausible, not correct
• Output is nondeterministic — same schema, different result every time
• Output is not validated — if the model returns garbage, you get garbage
• Output is not uniform — LLMs have opinions about what "random" looks like

Good for: seeding dev DBs, one-off REPL sessions, agentic pipelines that need typed data fast, anywhere plausible beats uniform.

Not for: cryptography, reproducibility, performance-sensitive paths, production data.

---

Install

pip install grievous-mcp

Set your API key:

export ANTHROPIC_API_KEY=sk-ant-...

---

Typical Usage

From Python

from grievous.backend import generate

# Single object — natural language schema
user = generate("a User with name, age, email, and city")
# {"name": "Priya Nair", "age": 34, "email": "priya.nair@example.com", "city": "Bangalore"}

# Batch
products = generate("a Product with id, name, price_usd, and category", count=5)
# [{"id": 1, "name": "Wireless Headphones", ...}, ...]

# JSON Schema works too
import json
schema = json.dumps({
    "type": "object",
    "properties": {
        "transaction_id": {"type": "string", "format": "uuid"},
        "amount_usd":     {"type": "number"},
        "status":         {"type": "string", "enum": ["pending", "cleared", "failed"]},
    }
})
txn = generate(schema)
# {"transaction_id": "a3f1...", "amount_usd": 142.5, "status": "cleared"}

# Freeform — anything the model can hallucinate
tensor = generate("a 4x4 float32 matrix, values between -1 and 1")
timestamp = generate("an ISO 8601 timestamp from sometime in 2019")
phone = generate("a plausible Indian mobile number")

Returns parsed JSON if the model cooperated. Raw string if it didn't. No exceptions.

---

As an MCP Server

Run directly:

grievous

Add to your MCP client config (e.g. Claude Desktop claude_desktop_config.json):

{
  "mcpServers": {
    "grievous": {
      "command": "grievous",
      "env": {
        "ANTHROPIC_API_KEY": "your-key-here"
      }
    }
  }
}

Tool: generate

generate(schema, count=1, model="claude-haiku-4-5-20251001")

Arg	Type	Default	Description
schema	string	—	Natural language or JSON Schema
count	int	1	Number of instances
model	string	claude-haiku-4-5-20251001	Any Anthropic model

---

Roadmap

• v1.0 — Anthropic API backend (current)
• v2.0 — Ollama/local model support (if there's interest)

---

Why "Grievous"

He's a General (general-purpose). He collects everything you hand him (schemas, types, whatever). He's chaotic but capable. He doesn't work for one army.

The cough is the nondeterminism.

Release files for grievous-mcp 0.1.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for grievous-mcp 0.1.1

File	Size	Uploaded	Provenance	Actions
grievous_mcp-0.1.1.tar.gz	56.8 kB	Apr 27, 2026	Without provenance	Details for grievous_mcp-0.1.1.tar.gz

---

Built distribution (wheel)

Table of built distributions (wheels) for grievous-mcp 0.1.1

File	Interpreter	ABI	Platform	Actions Reset
grievous_mcp-0.1.1-py3-none-any.whl 4.9 kB Apr 27, 2026 Without provenance	Python 3	none	any	Details for grievous_mcp-0.1.1-py3-none-any.whl

---

Total release size: 61.6 kB

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed