--- title: "Install" description: "Install QWED with pip, npm, Go, Cargo or Docker, then run qwed init to check the engines and connect a model provider." url: https://qwedai.tech/install --- # Install QWED runs in your own environment. Install the library for your language, then run qwed init to check the engines and connect a model provider. `$ pip install qwed` Python 3.10+ `$ npm install @qwed-ai/sdk` TypeScript `$ go get github.com/QWED-AI/qwed-verification/sdk-go/v7` Go `$ cargo add qwed` Rust `$ docker pull qwedai/qwed-verification` Container ## Python Python 3.10 or newer. One command installs the core engines (Math, Logic with Z3, SQL with SQLGlot, Code, Schema, Fact and Stats) and the `qwed` command line tool. shell ``` $ pip install qwed ``` ### Optional extras - `qwed[server]` The FastAPI server and its dependencies - `qwed[symbolic]` CrossHair, for the Symbolic engine - `qwed[pii]` PII masking with Microsoft Presidio - `qwed[langchain], qwed[crewai], qwed[llamaindex]` Framework integrations (experimental, not covered by CI) shell ``` $ pip install "qwed[server]" ``` ## Other languages and runtimes - ### TypeScript ``` $ npm install @qwed-ai/sdk ``` Latest on npm: 7.3.0. - ### Go ``` $ go get github.com/QWED-AI/qwed-verification/sdk-go/v7 ``` Releases are tagged sdk-go/vX.Y.Z; the module path ends in /v7. - ### Rust ``` $ cargo add qwed ``` - ### Docker ``` $ docker pull qwedai/qwed-verification ``` The API server. It needs API\_KEY\_SECRET and QWED\_CORS\_ORIGINS. [Self-hosting guide](https://docs.qwedai.com/advanced/self-hosting) ### From source shell ``` $ git clone https://github.com/QWED-AI/qwed-verification.git $ cd qwed-verification $ pip install -e ".[server,dev]" ``` ## First run Three commands take you from a fresh install to a working setup. - `qwed init` Checks the engines, sets up a model provider and creates a local API key. - `qwed doctor` Health check for the engines, provider, server and database. - `qwed test` Runs twelve deterministic checks. Needs no model and no server. qwed init, shortened ``` $ qwed init [QWED] Initializing verification engines... [ok] SymPy math engine ready [ok] Z3 logic engine ready [ok] AST code engine ready [ok] SQLGlot sql engine ready Running verification suite... [ok] 2+2=5 -> BLOCKED [ok] x>5 AND x<3 -> UNVERIFIABLE [ok] SELECT * WHERE 1=1 -> BLOCKED [ok] eval(user_input) -> VERIFIED (unsafe) All engines verified. QWED is operational. Step 1/3: LLM Provider Setup Step 2/3: API Key Step 3/3: Generate QWED API Key Starting local server... Warning: Save this key. It is shown only once. QWED is ready. ``` The suite feeds each engine a known-bad input, and every case must come back denied. This is the CLI’s output today, and two of its labels are out of date: `2+2=5` was checked and proven false, which the protocol records as `VERIFIED · DENY` rather than `BLOCKED`, and “All engines verified” means the engines are ready. Both are being fixed in [qwed-verification#438](https://github.com/QWED-AI/qwed-verification/issues/438). ## Model providers QWED uses a model only to translate natural language into a formal statement, and treats it as an untrusted translator. Pick a provider with `--provider`. Any OpenAI-compatible API works through `custom`. - `nvidia` - `openai` - `anthropic` - `gemini` - `custom` To run against a local model such as Ollama, see [LLM configuration](https://docs.qwedai.com/getting-started/llm-configuration) in the docs. shell ``` $ qwed init --provider anthropic # In CI, skip the prompts $ qwed init --non-interactive --provider nvidia ``` **Your data stays where you run QWED.** QWED does not send data anywhere by itself. Checks that use a model send the prompt to the provider you configure; with a self-hosted model, nothing leaves your infrastructure. Optional PII masking (`qwed[pii]`) redacts card numbers, social security numbers and email addresses before the model call. ## In CI with GitHub Actions Scan for secrets, unsafe code or unsafe shell commands on every push. With `output_format: sarif` the action writes `qwed-results.sarif`; upload it with [codeql-action/upload-sarif](https://github.com/github/codeql-action) to see findings in the Security tab. For pull request checks without writing a workflow, install the [QWED Security](https://qwedai.tech/platform/security) GitHub App. .github/workflows/qwed.yml ``` - uses: QWED-AI/qwed-verification@v7.3.0 with: action: scan-secrets # or scan-code, verify-shell paths: "**/*.py,**/*.env" output_format: sarif ``` ## Domain packages Each domain package adds checks for one field and installs on its own. - [qwed-finance](https://finance.qwedai.com) Banking, payments and financial calculations v3.0.1 `pip install qwed-finance` - [qwed-legal](https://github.com/QWED-AI/qwed-legal) Deadlines, amounts and legal claims v0.5.1 `pip install qwed-legal` - [qwed-tax](https://github.com/QWED-AI/qwed-tax) Tax compliance and withholding v0.2.1 `pip install qwed-tax` - [qwed-infra](https://github.com/QWED-AI/qwed-infra) Terraform, IAM and cloud cost v0.3.0 `pip install qwed-infra` - [qwed-open-responses](https://github.com/QWED-AI/qwed-open-responses) Agent tool calls, checked before they run v0.6.1 `pip install qwed-open-responses` - [qwed-mcp](https://github.com/QWED-AI/qwed-mcp) MCP server for any MCP client v0.2.2 `pip install qwed-mcp` - [qwed-a2a](https://github.com/QWED-AI/qwed-a2a) Agent-to-agent verification v0.3.0 `pip install qwed-a2a` - [qwed-ucp](https://github.com/QWED-AI/qwed-ucp) Universal Commerce Protocol transactions v0.3.0 `pip install qwed-ucp` [Next: the quickstart](https://docs.qwedai.com/getting-started/quickstart) [How the engines work](https://qwedai.tech/engines) [Install problem? Open an issue](https://github.com/QWED-AI/qwed-verification/issues/new/choose)