# Wyrd > Wyrd is open-source AI assurance and verification infrastructure. Verify AI systems before and after release, with verdicts you can trace, reproduce, and audit. ## Start here Declare -> Observe -> Verify an application you own, in this order. A connected Wyrd MCP server teaches the same route through `wyrd.guide {"topic":"start"}`. - [Set up Wyrd](/wyrd/get-started/setup/): run a local server you keep running, and connect MCP hosts. - [Verify your first agent](/wyrd/get-started/quickstart/): declare, observe, and verify the support-desk Service. - [Error catalog](/wyrd/api/errors/): every stable error code and its remediation. ## Home - [Wyrd docs](/wyrd/): the docs home, with a link to each published section. ## Get started - [What Wyrd verifies](/wyrd/get-started/introduction/): Wyrd checks that an AI system behaves the way its declaration says it should, before and after deployment. - [Set up Wyrd](/wyrd/get-started/setup/): Run a Wyrd server on your machine, save its first credential, and make an authenticated request. - [Verify your first agent](/wyrd/get-started/quickstart/): Run the support-desk API against your local server, get verdicts from three Verifiers, and query the evidence behind them. ## Build and verify - [Define a Service](/wyrd/build-and-verify/declare/define-a-service/): Declare a Service as a folder of Card files, with exact component versions and a table for the rows it records, then validate it locally. - [Bind a Verifier](/wyrd/build-and-verify/declare/bind-a-verifier/): Attach a Verifier to a Service or one of its components with verified_by, and choose whether your code or a schedule runs it. - [Declare and inspect a Service](/wyrd/build-and-verify/declare/declare-and-inspect/): Declare a Service and every Card it references in one call, then read back its exact versions and the relationships the server derived. - [Record a run and observations](/wyrd/build-and-verify/observe/record-a-run-and-observations/): Pull your Service's bundle, record rows inside a run, verify or submit them to a Verifier, and read them back by run ID. - [Write and query custom data with Bifrost](/wyrd/build-and-verify/observe/write-and-query-custom-data-with-bifrost/): Connect to Bifrost directly to register a table, load rows, and query any table you can read with SQL. - [Author a Verifier](/wyrd/build-and-verify/verify/author-a-verifier/): Write a Verifier Card from its inputs and requirements, starting with a one-line assertion and working up to a model judge that cites a document. - [Verify a request in real time](/wyrd/build-and-verify/verify/verify-a-request-in-real-time/): Call a bound Verifier from your service while a request is in flight, act on the verdict, and read the requirement breakdown. - [Monitor drift](/wyrd/build-and-verify/verify/monitor-drift/): Compare the inputs your service sees with a healthy baseline on a schedule, and get paged when they shift. - [Run a verification and inspect its evidence](/wyrd/build-and-verify/verify/run-a-verification-and-inspect-its-evidence/): Queue a Verifier with submit, follow its execution from any process, tell a failed verdict from a check that never ran, and read the stored evidence. - [Respond to a failed verification](/wyrd/build-and-verify/verify/respond-to-a-failed-verification/): Page someone when a submitted or scheduled Verifier fails, with an Operator Card, an operator connection, and on_failure, then check that the page was delivered. ## Connect - [Set up the LLM gateway](/wyrd/connect/set-up-the-llm-gateway/): Store a provider key with the Wyrd server, name a deployment for it, and call the model through Wyrd with your Wyrd credential. - [Make gateway requests from a framework](/wyrd/connect/call-wyrd-from-a-stock-client/): Run inference through a named Wyrd gateway deployment from the OpenAI, Anthropic, and Google GenAI SDKs, the OpenAI Agents SDK, and Pydantic AI. - [Connect an agent through MCP](/wyrd/connect/connect-an-agent-through-mcp/): Give Claude Code, Codex, Cursor, Copilot, Pi, or Hermes Agent Wyrd's tools with wyrd mcp install, against a local or deployed server. ## Operate - [Deploy Wyrd for a team](/wyrd/operate/deploy-wyrd-for-a-team/): Prepare PostgreSQL with a platform login and a tenant login, migrate it, start the server, and create the first tenant. - [Grant and inspect permissions](/wyrd/operate/grant-and-inspect-permissions/): See which roles a principal holds, grant or revoke a role, cut a principal off, and confirm the result in the audit log. - [Configure the verification runtime](/wyrd/operate/configure-verification-runtime/): Install the Python worker, choose which processes run Verifiers, size the worker pool, and diagnose startup refusals. - [Troubleshoot a deployment](/wyrd/operate/troubleshoot-a-deployment/): Find the cause of a Wyrd server that refuses to migrate or start, from the message it prints. ## Understand - [Cards, versions, and relationships](/wyrd/understand/cards-versions-and-relationships/): A Service declaration is a set of versioned Cards, each one a defined component of your system. - [How verification works](/wyrd/understand/how-verification-works/): Inputs, steps, requirements, and decide; verify versus submit; executions, verdicts, results, and Operators. - [Runs, observations, and retained evidence](/wyrd/understand/runs-observations-and-evidence/): How Wyrd records what a declared service actually does, and ties each observation to the version that ran. - [Bifrost](/wyrd/understand/bifrost/): Wyrd's analytical warehouse, the two ways to write to it, and when a written row becomes queryable. - [Identity and authorization](/wyrd/understand/identity-and-authorization/): Every call to Wyrd is made as a principal whose roles decide what it may do, and every decision is audited. ## Reference - [Verifier Card](/wyrd/reference/verifier-card/): Every field of a Verifier spec, its input slots, steps, checks, operators, references, citations, and outcomes. - [MCP tools](/wyrd/reference/mcp-tools/): Exact Wyrd tool names, permissions, argument and result shapes, refusals, and query limits. - [Extraction output](/wyrd/reference/extraction-output/): The exact JSON an extract step returns for every media format, how to address it from assertions and citations, and what extraction does not do. - [Verification runtime limits](/wyrd/reference/verification-runtime-limits/): Server environment settings that bound Verifier files, extraction, workspaces, Python code, and real-time deadlines, with their defaults and startup rules. - [Agent tools and authorization](/wyrd/reference/agent-tools-and-authorization/): The built-in tools a Verifier's Agent can call, how they are selected, and the permissions every call needs. - [Error codes](/wyrd/api/errors/): Every stable Wyrd error code with its HTTP status, title, and fix. - [HTTP API](/wyrd/api/openapi/): Where a Wyrd server serves its OpenAPI document and how to use it. ## Machine-readable reference - [llms-full.txt](/wyrd/llms-full.txt): every page below in full, with its embedded YAML and code.