Case ContextCopy Anchor Link

When you ask Claira to review a document, the AI only sees the text of that document. It does not know what your matter is about, who the key players are, or what makes a document relevant. Case Context fills that gap.

Case Context is a detailed, structured briefing on your matter that lives inside Claira. It opens with a short summary of why the review exists, then keeps every material fact from the documents you build it from. It gives the AI the background it needs to generate smarter, more accurate prompts and to interpret documents in the right context.

What Case Context doesCopy Anchor Link

• Informs AI prompts. When you use Claira's prompt generator, it draws on the approved context you pick to produce prompts tailored to your specific matter. Use with context on any Quickstart template fills that template in from it.
• Keeps your team aligned. Because contexts are shared across all users on a case, everyone works from the same understanding of the matter.
• Lives per case and organization. Each case within your organization gets its own contexts. They are scoped to that combination, so different cases have different contexts.

From documents to promptsCopy Anchor Link

Every review prompt built from a context passes through two steps, and Claira holds each one to a rule so that no detail is lost along the way:

1. From documents to Case Context, Claira keeps everything material. It removes repetition but never summarizes: a long, detailed document gives a long, detailed context. After each update, it checks that every fact the context already held is still there, and puts back any line the update left out. Each line it writes from a document ends with that document's number in the source documents list, so you can see where every fact came from.
2. From Case Context to prompt, the Prompt Generator selects but never shortens. It leaves out what your task does not need (a relevance prompt does not carry the privilege rules) and carries every fact the task does need in full: every name and its variants, dates, amounts, identifiers and issues. Before the prompt reaches your prompt box, Claira checks that every matter fact the task uses is in it.

Prompt History then records which version of which context a prompt was built from, and each bulk scan records which version of the prompt it ran, so you can trace any scan back to its source documents.

Multiple contexts per caseCopy Anchor Link

A case is not always a single workstream. The same document collection might serve the main litigation, a regulator's inquiry, and an internal investigation, each with its own people, questions, and sensitivities. Claira lets you keep a separate context for each workstream (up to 10 per case).

Each context has its own name, terminology, sections, source documents, and approval. One context is marked as the case default: the prompt generator preselects it, and the Agent reads and updates it in Agent Mode, unless you open a chat about another context with Revise with Agent.

Templates set the terminologyCopy Anchor Link

You do not have to be a litigator for Case Context to speak your language. When you create a context, you pick a template that sets what it is called and which sections it starts with:

Every template starts with three or four sections, and the sections are yours to shape: rename them, remove them, or add your own (up to 12 per context) after the context is created. The AI generation and the downloaded PDF and Word files follow whatever sections the context carries.

What Case Context is NOTCopy Anchor Link

It is worth being clear about boundaries:

• Not a case memo. A context records the facts your documents and your team state, in full. It is not your analysis of them.
• Not a replacement for case strategy. Strategic decisions about how to handle the matter belong in your case management system, not here.
• Not a place for privileged strategy commentary. Contexts are visible to all users on the case. Do not include work product, mental impressions, or litigation strategy.

When to use Case ContextCopy Anchor Link

Case Context is most valuable when:

• The matter involves litigation, an arbitration, an employment dispute, a regulatory inquiry, or an investigation where relevance criteria and sensitivities are well defined.
• Multiple reviewers need consistency: a shared context ensures the AI generates the same quality of prompts for everyone.
• You want AI-tailored prompts rather than generic ones. The prompt generator uses the context you pick to produce prompts specific to your matter.

For simple or ad hoc review tasks, you can use Claira without Case Context. But for any matter where accuracy, consistency, and defensibility matter, setting up at least one context is strongly recommended.

How it works technicallyCopy Anchor Link

Each context is stored as structured data (JSON) and rendered as human-readable sections in the Claira interface. Each section carries instructions that tell Claira what to capture there when it builds the context from documents, and you can edit them while the context is a draft. A draft can also be revised by the Agent in a chat about that context, which saves changes only after you confirm. When the AI generates a prompt, it reads the sections of the context you selected, under their own names, to understand the matter before producing its output. It does not read the summary or the source numbers.

When Claira builds a context from documents, it reads each document's extracted text and, when it fits, the document's own file (a PDF, image, audio or video file). Scanned documents without OCR, recordings and videos can therefore feed a context too. See What Claira reads from each document.

The generated prompt must be self-contained: it includes the matter details needed for review directly in prompt text. During document scans, the model receives the prompt plus document/evidence content; it does not receive Case Context as a separate hidden payload on every scan.

---

Need help? Contact support@claira.to