Inferwall
inferwall.com
- Free plan
- Windows
- macOS
- Linux
- API
What Inferwall is
Inferwall is a self-hosted AI firewall that scans LLM application inputs and outputs for threats such as prompt injection, jailbreaks, data leakage, and unsafe content. It combines heuristic, machine-learning, semantic, and configurable LLM-judge detection layers, and can run as a Python SDK, standalone HTTP server, or transparent reverse proxy. Listed first-party wrappers include OpenAI, Anthropic, LangChain, and FastAPI; other supported integrations include Ollama, Hugging Face, LlamaIndex, Pydantic, vLLM, and Mistral. Teams can set allow, flag, or block thresholds by route and add or override signatures and version policies under ~/.inferwall/. The Lite profile is about 8 MB and requires no GPU; Standard adds about 730 MB of models, while Full adds the LLM-judge layer and requires LLM credentials. The site says the Rust heuristic engine runs in under 0.3 ms at p99. The engine uses Apache 2.0, while community signatures use CC BY-SA 4.0. Inferwall says relevant signatures map to MITRE ATLAS techniques, but cautions that no detection system is perfect and presents the tool as one layer of a defense-in-depth approach.
Who it suits
Inferwall suits teams running LLM applications that want a self-hosted layer for scanning inputs and outputs. Its deployment choices, integrations, and route-level policies may fit teams that need to tailor how threats are handled.
Good to know, and what to watch
- Three deployment options are listed.
- Supports allow, flag, or block thresholds per route.
- Rules can be added or overridden.
- Lite profile needs no GPU.
- Supports listed wrappers and ecosystem integrations.
- No detection system is perfect.
- Full profile requires LLM credentials.
- Standard adds about 730 MB of models.
The verdict
Inferwall offers configurable scanning with several deployment approaches and integration options. Its own caveat matters: it should be treated as one defense layer, not a guarantee that threats will be detected.
What the maker publishes
- Additional ecosystem support
- The site lists Ollama, Hugging Face, LlamaIndex, Pydantic, vLLM, and Mistral among supported stack integrations. source
- Caveat
- The site says no detection system is perfect and describes Inferwall as one layer in a defense-in-depth strategy. source
- Deployment
- Inferwall can be deployed as a Python SDK, standalone HTTP server, or transparent reverse proxy. source
- Deployment
- It can run as a Python SDK, standalone HTTP server, or reverse proxy. source
- Detection layers
- It combines heuristic, machine learning, semantic, and configurable LLM judge detection layers. source
- Detection layers
- It combines heuristic, ML classifier, semantic, and configurable LLM-judge detection layers. source
- Install profiles
- The site describes Lite as about 8 MB, Standard as including about 730 MB of models, and Full as including those models plus LLM credentials. source
- Integrations
- The site lists first party wrappers for OpenAI, Anthropic, LangChain, and FastAPI. source
- Integrations
- The site lists first-party wrappers for OpenAI, Anthropic, LangChain, and FastAPI. source
- License
- The engine is Apache 2.0 and community signatures are CC BY-SA 4.0. source
- Licensing
- The engine is licensed under Apache 2.0 and community signatures under CC BY-SA 4.0. source
- Limitations
- The site states that no detection system is perfect and presents Inferwall as one layer of a defense in depth strategy. source
- Performance
- The site says its Rust heuristic engine runs in under 0.3 ms at p99 and requires no GPU in the Lite profile. source
- Performance
- The site says its Rust heuristic engine runs in under 0.3 ms at p99 and the Lite profile requires no GPU. source
- Policy controls
- Users can set allow, flag, or block thresholds per route and extend or override signatures under ~/.inferwall/. source
- Profiles
- The listed Lite profile is about 8 MB; Standard adds about 730 MB of models; Full adds the LLM-judge layer and requires LLM credentials. source
- Purpose
- Inferwall describes itself as an AI firewall that scans LLM app inputs and outputs against signatures for threats including prompt injection, jailbreaks, data leakage, and unsafe content. source
- Purpose
- Inferwall scans LLM app inputs and outputs for prompt injection, jailbreaks, data leakage, content safety issues, and agentic threats. source
- Rules
- Users can override shipped signatures, add their own, and version policies under ~/.inferwall/. source
- Security mapping
- The site says its prompt injection, jailbreak, prompt extraction, and data leakage signatures map to MITRE ATLAS techniques AML.T0051, AML.T0054, AML.T0056, and AML.T0057. source
Checked against the maker's pages on 7 Oct 2026.