# Nora > AI that catches its own mistakes. Nora watches every answer your agents give in production, finds the root cause when one is wrong, and fixes it. ## Docs - [Analysis sessions](https://docs.platform.nora.my/build/causal-graph/analysis.md): Interactive reasoning over the graph — hypothesize, test, iterate. - [Attach a graph to an Agent](https://docs.platform.nora.my/build/causal-graph/attach.md): Give an Agent structured domain knowledge as a first-class data source. - [Edges](https://docs.platform.nora.my/build/causal-graph/edges.md): The causal relationships between variables — direction, strength, and mechanism. - [Extract from documents](https://docs.platform.nora.my/build/causal-graph/extraction.md): Bootstrap a causal graph from your existing knowledge — proposals you review and accept. - [Causal graph overview](https://docs.platform.nora.my/build/causal-graph/overview.md): A visual map of cause and effect in your domain — used to explain, verify, and analyze. - [Variables](https://docs.platform.nora.my/build/causal-graph/variables.md): The nodes of your causal graph — the things you measure. - [Verify agent answers](https://docs.platform.nora.my/build/causal-graph/verification.md): Cross-check every agent output against your causal graph before it reaches the user. - [Chunk & parse](https://docs.platform.nora.my/build/foundry/chunk.md): Break documents into retrievable pieces of the right size. - [Clean stage](https://docs.platform.nora.my/build/foundry/clean.md): Redact PII, deduplicate, and remove noise before your data hits the index. - [Connectors](https://docs.platform.nora.my/build/foundry/connectors.md): Reusable, credentialed connections to the outside world. - [Enrichment](https://docs.platform.nora.my/build/foundry/enrich.md): Add embeddings, tags, entity extractions, and language labels to your chunks. - [Source blocks](https://docs.platform.nora.my/build/foundry/ingest.md): Pull raw data in from cloud drives, databases, APIs, and files. - [Run logs](https://docs.platform.nora.my/build/foundry/logs.md): Inspect any pipeline run — stage by stage, item by item. - [Output stage](https://docs.platform.nora.my/build/foundry/output.md): Persist enriched chunks into the store your Agents read from. - [Foundry overview](https://docs.platform.nora.my/build/foundry/overview.md): Turn raw data into knowledge your agents can use. A visual pipeline for ingest, clean, chunk, enrich, and index. - [Triggers](https://docs.platform.nora.my/build/foundry/triggers.md): How Foundry pipelines start — manual, scheduled, watch, or webhook. - [Attach knowledge to agents](https://docs.platform.nora.my/build/knowledge/attach.md): Give each Agent access to exactly the knowledge it needs — no more, no less. - [Knowledge overview](https://docs.platform.nora.my/build/knowledge/overview.md): The documents and structured data your agents answer from. - [Versioning & supersede](https://docs.platform.nora.my/build/knowledge/supersede.md): Retire outdated knowledge without losing the audit trail. - [Tag & organize](https://docs.platform.nora.my/build/knowledge/tag.md): Group documents by department, audience, feature, or freshness so agents pull the right subset. - [Upload documents](https://docs.platform.nora.my/build/knowledge/upload.md): Add files to your knowledge — drag & drop, paste URLs, or import from a link. - [Forgetting](https://docs.platform.nora.my/build/memory/forgetting.md): Nothing is permanent unless you say so — TTLs, decay, and safe supersede. - [Relationship graphs](https://docs.platform.nora.my/build/memory/graphs.md): Store structured facts as entity-relationship triples the Agent can traverse. - [Notes & recall](https://docs.platform.nora.my/build/memory/notes.md): The Agent writes notes as it runs, and reads them back when relevant. - [Memory overview](https://docs.platform.nora.my/build/memory/overview.md): What agents remember — from a single session up to persistent knowledge learned over time. - [Procedural rules](https://docs.platform.nora.my/build/memory/procedural.md): Standing behaviors — 'always do X before Y', 'never do Z without approval'. - [Scoping](https://docs.platform.nora.my/build/memory/scoping.md): Keep memory isolated per user, per tenant, per conversation — without duplicating data. - [Memory spaces](https://docs.platform.nora.my/build/memory/spaces.md): Pick the right container for what you want to remember. - [Retrieval diagnostics](https://docs.platform.nora.my/build/retrieval/diagnostics.md): Why isn't my chunk showing up? - [Filters](https://docs.platform.nora.my/build/retrieval/filters.md): Narrow retrieval candidates by metadata — tag, source, date, entity. - [Grounding](https://docs.platform.nora.my/build/retrieval/grounding.md): Verify retrieved facts against a knowledge graph before returning them. - [Hybrid search](https://docs.platform.nora.my/build/retrieval/hybrid-search.md): Combine semantic similarity with keyword matching — the best of both. - [Retrieval overview](https://docs.platform.nora.my/build/retrieval/overview.md): How agents find the right context at inference — hybrid ranking, filters, and grounding. - [Retrieval presets](https://docs.platform.nora.my/build/retrieval/presets.md): Save a retrieval configuration and reuse it across Agents. - [Ranking & weights](https://docs.platform.nora.my/build/retrieval/ranking.md): The knobs that decide which chunks come back first. - [Action block](https://docs.platform.nora.my/build/workflow/action-block.md): Side-effects that fire after the Agent decides — no LLM picking arguments. - [Agent block](https://docs.platform.nora.my/build/workflow/agent-block.md): The reasoning core of every Flow — a role, a model, tools, and memory. - [Flow canvas](https://docs.platform.nora.my/build/workflow/canvas.md): The infinite workspace where you drag, wire, and arrange blocks. - [Workflow overview](https://docs.platform.nora.my/build/workflow/overview.md): The Flow canvas is where you build an agent. Drag blocks in, wire them together, and ship. - [Publishing](https://docs.platform.nora.my/build/workflow/publish.md): Ship the Draft as an immutable version — the one production traffic sees. - [Flow settings](https://docs.platform.nora.my/build/workflow/settings.md): Flow-wide knobs — limits, approvals, defaults, and access. - [Tool block](https://docs.platform.nora.my/build/workflow/tool-block.md): Give your Agent the ability to call APIs, run code, or query databases. - [Trigger block](https://docs.platform.nora.my/build/workflow/trigger-block.md): How Flows start — chat, webhook, schedule, or event. - [Wiring blocks](https://docs.platform.nora.my/build/workflow/wiring.md): Connect blocks so data flows through them, and branch on the Agent's decisions. - [Changelog](https://docs.platform.nora.my/changelog.md): What's new in Nora. - [nora auth](https://docs.platform.nora.my/cli/auth.md): Sign in, switch workspaces, manage tokens. - [nora flows](https://docs.platform.nora.my/cli/flows.md): Manage Flows from the terminal — list, edit, publish, roll back. - [Install the CLI](https://docs.platform.nora.my/cli/install.md): Get the nora command on your machine and sign in. - [Command reference](https://docs.platform.nora.my/cli/reference.md): Every nora subcommand at a glance. - [Agents & Blocks](https://docs.platform.nora.my/concepts/agents.md): The five kinds of Block, and how they connect. - [Flows](https://docs.platform.nora.my/concepts/flows.md): One canvas, one agent workflow — versioned and safe to publish. - [Memory](https://docs.platform.nora.my/concepts/memory.md): What your agent remembers between runs — and what you control. - [Core concepts](https://docs.platform.nora.my/concepts/overview.md): Workspaces, Flows, and Blocks — the vocabulary you'll use everywhere in Nora. - [Workspaces](https://docs.platform.nora.my/concepts/workspaces.md): Isolated containers for your Flows, data, and teammates. - [Google Workspace](https://docs.platform.nora.my/integrations/google-workspace.md): Use Drive, Gmail, and Calendar as data sources and action targets. - [MCP — coding agents](https://docs.platform.nora.my/integrations/mcp.md): Drive Nora from Claude Desktop, Cursor, or any MCP-capable coding agent. - [Integrations](https://docs.platform.nora.my/integrations/overview.md): Connect Nora to the tools you already use. - [Welcome to Nora](https://docs.platform.nora.my/introduction.md): AI that catches its own mistakes. - [Quickstart](https://docs.platform.nora.my/quickstart.md): Sign up, build your first agent, and get an answer in under 10 minutes. - [Annotate examples](https://docs.platform.nora.my/reliable/datasets/annotate.md): Add expected outputs, rubrics, or verdicts so simulation can grade objectively. - [Coverage](https://docs.platform.nora.my/reliable/datasets/coverage.md): Is your dataset representative of what your Agents actually see? - [Create a dataset](https://docs.platform.nora.my/reliable/datasets/create.md): Four ways to build one — from traces, from clusters, from files, from scratch. - [Holdout sets](https://docs.platform.nora.my/reliable/datasets/holdout.md): Test data the Improvement flow never sees — the honest measure. - [Datasets overview](https://docs.platform.nora.my/reliable/datasets/overview.md): Curated collections of examples — the fuel for simulation, tests, and improvement. - [Dataset versioning](https://docs.platform.nora.my/reliable/datasets/versioning.md): Datasets change. Track versions so simulation results stay comparable. - [Approve & deploy](https://docs.platform.nora.my/reliable/optimization/approve.md): Human gate before the change goes live — with routing, sign-off, and canary rollout. - [Diagnose a failure](https://docs.platform.nora.my/reliable/optimization/diagnose.md): Locate the specific step in the Flow that produced the wrong answer. - [Improvement history](https://docs.platform.nora.my/reliable/optimization/history.md): Every improvement — what it fixed, whether it worked, and whether it lasted. - [Which lever to tune](https://docs.platform.nora.my/reliable/optimization/levers.md): Prompt, tool, retrieval, memory, rules — which change makes sense for this failure? - [Optimization overview](https://docs.platform.nora.my/reliable/optimization/overview.md): Turn confirmed signals into shippable improvements — proposals, simulation, approval, deploy. - [Improvement proposals](https://docs.platform.nora.my/reliable/optimization/proposals.md): Candidate fixes generated from the root cause — with diffs, predictions, and choices. - [Clusters](https://docs.platform.nora.my/reliable/signals/clusters.md): Signals that share a root cause — fix one, fix many. - [Confirm or mute](https://docs.platform.nora.my/reliable/signals/confirm.md): Triage a signal in seconds — is this real or noise? - [From signal to fix](https://docs.platform.nora.my/reliable/signals/from-signal-to-fix.md): The full journey — from a single 👎 to a deployed improvement, live. - [Signals overview](https://docs.platform.nora.my/reliable/signals/overview.md): Every bad answer, missed retrieval, or user complaint — surfaced, clustered, and ready to fix. - [Regressions](https://docs.platform.nora.my/reliable/signals/regressions.md): A previously-fixed problem came back. Catch it fast. - [Where signals come from](https://docs.platform.nora.my/reliable/signals/sources.md): The channels that feed the queue — feedback, errors, verification, and more. - [Baselines](https://docs.platform.nora.my/reliable/simulation/baselines.md): Pin a version as the reference — every change is measured against it. - [Before/after](https://docs.platform.nora.my/reliable/simulation/before-after.md): Side-by-side comparison — did the change actually improve things? - [Simulation overview](https://docs.platform.nora.my/reliable/simulation/overview.md): Run agent changes against your data before they hit production. - [Cost vs. quality](https://docs.platform.nora.my/reliable/simulation/pareto.md): The tradeoff frontier — cheaper isn't always worse; more expensive isn't always better. - [Regression checks](https://docs.platform.nora.my/reliable/simulation/regressions.md): Automatic detection of examples that used to pass and now don't. - [Run a simulation](https://docs.platform.nora.my/reliable/simulation/run.md): Kick one off from the app or CLI — pick the target, pick the dataset, go. - [Cost & latency](https://docs.platform.nora.my/reliable/traces/cost.md): Per-step cost and time — spot where the money and milliseconds go. - [Feedback on a trace](https://docs.platform.nora.my/reliable/traces/feedback.md): Thumbs, corrections, and notes — added right where the answer went wrong. - [Grounding & citations](https://docs.platform.nora.my/reliable/traces/grounding.md): Which parts of an answer are backed by evidence, and where the evidence came from. - [Inspect a run](https://docs.platform.nora.my/reliable/traces/inspect.md): The main Trace view — a timeline of everything that happened. - [Traces overview](https://docs.platform.nora.my/reliable/traces/overview.md): Every run of every agent, recorded end-to-end. Inspect exactly what happened. - [Search & filter](https://docs.platform.nora.my/reliable/traces/search.md): Find any trace — by content, by cost, by outcome, by anything. - [Tool calls](https://docs.platform.nora.my/reliable/traces/tool-calls.md): Everything the Agent asked its tools to do, and what came back. - [Audit trail](https://docs.platform.nora.my/reliable/versions/audit.md): Every publish, rollback, restore, and edit — logged with actor, time, and reason. - [Version diff](https://docs.platform.nora.my/reliable/versions/diff.md): See exactly what changed between two versions — line by line, block by block. - [Drafts](https://docs.platform.nora.my/reliable/versions/drafts.md): Where you edit safely — never affects production until you publish. - [Versions overview](https://docs.platform.nora.my/reliable/versions/overview.md): Treat agent configurations as code. Frozen, auditable, instantly rollback-able. - [Publish a version](https://docs.platform.nora.my/reliable/versions/publish.md): Freeze the current Draft and point production traffic at it. - [Rollback](https://docs.platform.nora.my/reliable/versions/rollback.md): Point production traffic back to a prior version — instant, atomic, safe.