AI Detect
github.com
Spec sheet
- What it does
- AI Detect classifies text sentence by sentence for likely AI authorship on the user’s own machine.Read 15 Oct 2026 · github.com
- Purpose
- AI Detect is an open-source tool that scores writing sentence by sentence for likely AI-generated text.Read 7 Oct 2026 · github.com
- Audience
- The maker positions it for people checking commissioned or unpublished copy locally, including users handling batches of writing.Read 7 Oct 2026 · github.com
- Integrations
- The MCP server can be used with Claude Desktop, Claude Code, or another MCP client.Read 7 Oct 2026 · github.com
- Privacy
- Text inference runs on the user's hardware; the repository says the only network call is the initial model download.Read 7 Oct 2026 · github.com
- Support
- The repository says users give up a support contract and polished dashboard when choosing this local tool.Read 7 Oct 2026 · github.com
- License
- The project is released under the MIT license.Read 15 Oct 2026 · github.com
- Accuracy caveat
- The guide reports that short factual human sentences can be falsely flagged while long conversational sentences often pass.Read 15 Oct 2026 · houtini.com
- Benchmark claim
- Houtini says the DeBERTa detector ranked first on the RAID detection benchmark.Read 15 Oct 2026 · github.com
- Business classification
- Companies House classifies HOUTINI LIMITED under SIC 62090, other information technology service activities.Read 15 Oct 2026 · find-and-update.company-information.service.gov.uk
- Caveat
- The maker describes the software as beta and recommends treating its scores as a signal across a batch rather than a verdict on an individual sentence.Read 7 Oct 2026 · github.com
- Command-line interface
- The CLI accepts text or files, compares two versions, emits JSON, and lets users choose model and device.Read 15 Oct 2026 · github.com
- Company founding
- LinkedIn lists Houtini as founded in 2025.Read 15 Oct 2026 · linkedin.com
- Company headquarters
- LinkedIn lists Houtini’s headquarters as London.Read 15 Oct 2026 · linkedin.com
- Company size
- LinkedIn lists Houtini as having 2–10 employees.Read 15 Oct 2026 · linkedin.com
- Default model download
- The default model downloads about 1.7 GB from Hugging Face on first use and is then cached locally.Read 15 Oct 2026 · github.com
- Detection model
- The default detector uses desklib/ai-text-detector-v1.01, based on a 304-million-parameter DeBERTa-v3-large model.Read 15 Oct 2026 · github.com
- Diagnostics
- It flags writing patterns such as missing contractions, formal verbs, and formal transitions, and provides plain-English rewrite suggestions.Read 7 Oct 2026 · github.com
- Hardware
- A GPU is optional: the default is CPU, with CUDA used automatically when available.Read 7 Oct 2026 · github.com
- Hardware support
- CPU execution is the default, while CUDA is used automatically when an available GPU is detected.Read 15 Oct 2026 · github.com
- Independent deployment status
- Glama states that this MCP server cannot be deployed through its platform.Read 15 Oct 2026 · glama.ai
- Independent documentation assessment
- Glama’s page reproduces instructions for CLI commands, MCP configuration, model downloads and the five exposed tools.Read 15 Oct 2026 · glama.ai
- Independent listing
- Glama lists AI Detect as a Python, local MCP server for AI text detection.Read 15 Oct 2026 · glama.ai
- Input limit
- Sentences shorter than five words are skipped rather than scored.Read 7 Oct 2026 · github.com
- Installation requirement
- The accompanying setup guide lists Python 3.9 or newer, pip and roughly 3 GB of disk space for the first download.Read 15 Oct 2026 · houtini.com
- Interfaces
- It is distributed as a command-line tool, Python package, and MCP server.Read 7 Oct 2026 · github.com
- Legal entity
- UK Companies House lists the maker as HOUTINI LIMITED, company number 16683235, an active private limited company incorporated on 31 August 2025.Read 15 Oct 2026 · find-and-update.company-information.service.gov.uk
- Light model
- An optional RoBERTa-base int8 ONNX model is provided for smaller and faster scans and is about 126 MB.Read 15 Oct 2026 · github.com
- Limit: detector robustness
- The RAID paper reports that current detectors can be fooled by adversarial attacks, sampling changes, repetition penalties and unseen generative models.Read 15 Oct 2026 · arxiv.org
- Limit: missing SaaS features
- The repository says users give up a polished dashboard, team accounts and a support contract compared with hosted detectors.Read 15 Oct 2026 · github.com
- Limit: product maturity
- The repository describes AI Detect as beta software and warns users to expect rough edges.Read 15 Oct 2026 · github.com
- MCP integration
- AI Detect includes an MCP server that can be used by Claude Desktop, Claude Code and other MCP clients.Read 15 Oct 2026 · github.com
- MCP tools
- Its MCP server exposes text detection, file detection, text comparison, model-status and model-listing tools.Read 15 Oct 2026 · github.com
- Model developer interface
- The Hugging Face page provides Transformers pipeline and AutoTokenizer/AutoModel loading examples.Read 15 Oct 2026 · huggingface.co
- Model repository facts
- The Hugging Face model card labels desklib/ai-text-detector-v1.01 as an English text-classification model with an MIT license and a 1.75 GB repository.Read 15 Oct 2026 · huggingface.co
- Models
- Its default model is a 304-million-parameter DeBERTa-v3-large model, with a 126 MB RoBERTa ONNX option for faster triage.Read 7 Oct 2026 · github.com
- Notable news
- Houtini’s AI Detect guide was published on 1 March 2026 and updated on 2 October 2026.Read 15 Oct 2026 · houtini.com
- Observed speed
- The maker’s guide reports a 1,500-word scan taking roughly four to six minutes on CPU and about 40 seconds on an RTX 3080.Read 15 Oct 2026 · houtini.com
- Offline use
- After downloading and caching the model, it runs offline.Read 7 Oct 2026 · github.com
- Pattern diagnostics
- The tool flags signals such as missing contractions, formal verbs, formal transitions, vague abstractions and uniform sentence rhythm.Read 15 Oct 2026 · github.com
- Privacy behavior
- After downloading the model, inference runs locally and the text is not uploaded to a service.Read 15 Oct 2026 · github.com
- Public API
- The project documents an importable Python package and MCP interface but does not document a hosted HTTP API or SDK.Read 15 Oct 2026 · github.com
- Registered office
- Companies House gives HOUTINI LIMITED’s registered office as Timberly, South Street, Axminster, Devon, United Kingdom, EX13 5AD.Read 15 Oct 2026 · find-and-update.company-information.service.gov.uk
- Rewrite guidance
- Each detected pattern can include a plain-English suggestion for rewriting the sentence.Read 15 Oct 2026 · github.com
- Self-hosting
- The repository can be installed from source with pip and run as the ai-detect CLI or ai-detect-mcp server.Read 15 Oct 2026 · github.com
- Short-text handling
- Sentences shorter than five words are skipped instead of being classified.Read 15 Oct 2026 · github.com
- Storage
- The default model download is about 1.7 GB, while the light model download is about 126 MB.Read 7 Oct 2026 · github.com
Plans and prices
Free Free no API key · no usage limits · local inference · offline after model download Read 15 Oct 2026 · github.com
Where Specifiction places it
- AI Content Detection Software#6 of 459.0
- AI Detection Software#4 of 299.0