CLARA

AI-enabled product intelligence by DSTA

Product clarity, grounded in evidence.

CLARA reads your programme’s knowledge base, turns raw field notes into evidence-backed insights, and files them as research artefacts across the Research, Design, and Test phases of the ProductOps pipeline — personas, journeys, PRDs, storyboards, test plans — cited to source, never fabricated.

Cites every findingFiles back to your KBAir-gap ready

How she thinks

Read → Draft → File. Cited at every step.

CLARA runs a disciplined loop over your knowledge base. The discipline is the point: she won’t give you an answer she can’t back with a citation.

Read

CLARA traverses your programme's knowledge base via MCP and reads the source items. She doesn't search the web; she doesn't invent context. The corpus is the boundary.

Draft

Every finding points to the exact source it came from. If the evidence is thin, CLARA flags the gap and asks before continuing.

File

Artefacts file back into your knowledge base under a disciplined path. CLARA creates the hierarchy top-down on first use — no manual folder setup, no improvised locations.

Where CLARA earns her keep

Replace the grind. Keep the chain.

CLARA isn’t a chatbot. She’s your AI research assistant. She replaces the slow, error-prone parts of producing research artefacts across Research, Design, and Test phases of the ProductOps pipeline — and chains the outputs so your evidence stays intact from transcript to test plan.

Research at speed

CLARA turns raw interview transcripts and field observations into evidence-backed insights — themes, friction points, problem statement, success criteria — in a single cited pass. The kind of work that takes a research lead a week or two lands as a draft in minutes. Personas, journeys, PRDs — the rest of the chain drafts in minutes too. The hours saved go back where they belong: understanding real users.

Manual1–2 weeks
With CLARAminutes

Chained by design

The chain runs end-to-end. CLARA prepares for the field by searching past programmes for prior knowledge that shapes the interview guide. After fieldwork, the synthesis grounds the persona, the persona shapes the journey, and the journey extends through service blueprints, operational scenarios, mission threads, capability specs, and PRDs. Evidence preserved at every stage.

Prior knowledgeInterview guideFieldworkSynthesisPersonaJourneyPRD

Citations at source level

Every finding points to the exact source item it came from. If a stakeholder asks ‘where did this come from?’, you can answer instantly — the link is in the artefact.

Cross-programme referencing

CLARA's prior-knowledge step searches across all programmes — not just yours — to surface what's already been learned about a topic. Don't rediscover what another team already knows.

What CLARA’s skills do

Fifteen skills. Three phases of work.

CLARA’s catalogue of skills spans Research, Design, and Test — from initial discovery through product definition, storyboards, and test plans. Each skill drafts an artefact that files to a predictable path and feeds the next.

Research

Prior-knowledge summariser

Surface what past programmes already learned about a topic — before starting fresh research.

Files to

Knowledge Base/{track}/Prior-knowledge/{topic}
Research

Heuristic evaluator

Run a Nielsen heuristic usability evaluation of an existing or competitor product — from screenshots, a URL, or a deck — into a scored, evidence-anchored report.

Files to

Knowledge Base/{track}/Heuristic-evaluations/{product-name}
Research

Interview-guide generator

Generate a field-ready interview guide targeted at the data the team actually needs to surface.

Files to

Knowledge Base/{track}/Interview-guides/{topic}
Research

Research synthesiser

Turn interview transcripts and field notes into a synthesis of themes, friction, problem statement, and success criteria.

Files to

Knowledge Base/{track}/Research-synthesis
Research

Persona generator

Draft a persona from research evidence — every claim sourced.

Files to

Knowledge Base/{track}/Personas/{persona-name}
Research

Journey-map drafter

Draft a current-state journey for a persona — stages, touchpoints, emotions, friction, opportunities.

Files to

Knowledge Base/{track}/Journeys/{journey-scope}
Research

Service-blueprint drafter

Extend a journey into a service blueprint — making the back-stage systems, processes, and teams visible.

Files to

Knowledge Base/{track}/Service-blueprints/{journey-scope}
Research

Operational-scenario generator

Draft an operational scenario from operator research and a capability brief — operator, environment, decisions, failure modes.

Files to

Knowledge Base/{track}/Operational-scenarios/{scenario-title}
Research

Capability-spec generator

Turn an operational scenario into measurable capability requirements — functional, performance, environmental.

Files to

Knowledge Base/{track}/Capability-specs/{capability-name}
Research

Mission-thread mapper

Map the end-to-end mission thread — actors, systems, and data flows that produce an operational outcome.

Files to

Knowledge Base/{track}/Mission-threads/{mission-task}
Research

Problem-impact ranker

Rank captured problems by impact — reach, severity, evidence, and leverage — into a tiered queue that says which problem to build for first.

Files to

Knowledge Base/{track}/Problem-impact-analysis
Research

PRD generator

Draft a first-pass PRD from research synthesis and stakeholder context — clarifying questions where inputs are thin.

Files to

Knowledge Base/{track}/PRDs/{prd-title}
Design

Before/after journey mapper

Lay today's journey beside the future-state with the product — phase by phase, each tagged with the ranked problems it addresses.

Files to

Knowledge Base/{track}/Before-after-journeys/{journey-scope}
Design

Capability-storyboard scripter

Script a visual storyboard of a capability in use — beat-by-beat panels that turn an operational scenario into something operators can react to before any platform is acquired.

Files to

Knowledge Base/{track}/Capability-storyboards/{storyboard-title}
Test

Test-plan generator

Draft a complete test plan — objective, scenarios, participants, measurement, analysis, and validity risks — from a PRD and its success criteria.

Files to

Knowledge Base/{track}/Test-plans/{test-name}

For platform teams

Deploy CLARA in your stack.

A single Markdown file. Install once per environment. CLARA is vendor-neutral and air-gap clean — she runs against whichever KB-aware LLM your platform already provides.

Air-gap readyVendor-neutralZero per-programme setup

What you need

CLARA's skill file

A single Markdown file from CLARA's repo — either `dist/system-prompt.md` (flat) or `dist/SKILL.md` (skill-protocol wrapper). Pick the one that matches your stack.

Knowledge-base MCP

Already wired into your LLM. CLARA does not require a separate connector — she uses whichever knowledge-base MCP (Plane, Confluence, and the like) your stack provides.

Nothing else

No new infrastructure, no per-programme provisioning, no permissions request. Once installed, any team can invoke CLARA against their own knowledge base.

Choose your platform

One CLARA, two knowledge-base backends. Pick the build that matches where your programme’s knowledge base lives, then connect the matching MCP server.

CLARA for Plane

Access repo

For programmes whose knowledge base lives in Plane. Artefacts file back as nested project pages under the Knowledge Base hierarchy, cited by CLARA-assigned Session IDs.

Connect Plane MCP

Plane MCP must be connected manually — it isn't available as a one-click connector in Claude yet. Set up the Plane MCP server against your workspace by following Plane's step-by-step guide, then CLARA uses it.

Plane MCP setup guide

CLARA for Confluence

Access repo

For programmes whose knowledge base lives in Confluence. Artefacts file back as pages under the Knowledge Base hierarchy, with field notes tracked by CLARA's Session-ID scheme.

Connect Confluence MCP

You can find the Confluence (Atlassian) MCP and connect it directly in Claude — no manual server setup required. Add it from Claude's connectors and CLARA uses it. Reference the Atlassian MCP project if you need details.

Atlassian MCP (mcp-atlassian)

Choose your stack

The install is the same shape across every stack: load CLARA’s skill file into the LLM’s system context. The mechanism differs by platform.

LLM gateway

LiteLLM, custom API wrapper, internal chat backend.

Where the file goes

Prepend the file contents to the system message of every outbound LLM request.

  1. Download dist/system-prompt.md from CLARA's repo.

  2. Load it once at gateway startup; cache the contents in memory.

  3. Prepend the contents to the system role on every request before forwarding.

Example

// Pseudo-code for a gateway request handler.
const claraPreamble = await fs.readFile("clara/system-prompt.md", "utf8");

async function handle(req: ChatRequest) {
  const messages = [
    { role: "system", content: claraPreamble },
    ...req.messages,
  ];
  return forwardToLLM({ ...req, messages });
}

Verify the install

Open a fresh chat against the configured LLM and paste the invocation below. CLARA should respond with a one-line route confirmation and a batched question for the missing inputs.

Paste this

Use CLARA's `persona-generator` for <a programme with a knowledge base you can access>.

Expected response

CLARA confirms which artefact she’ll run and against which programme, then asks for the track, persona name, and whether to search your knowledge base or accept paste-in inputs — all in one message.

Keep it current

SHA-pinned

Each CLARA release stamps a git SHA into the skill file and downstream artefacts. Pin a specific SHA in your config; bump deliberately.

Rarely changes

CLARA's content updates infrequently — new artefacts, KB convention revisions, persona refinements. No constant churn to keep up with.

One-time per release

Each install is a one-time action per CLARA release. Bump the skill file in your environment when a new SHA ships.

CLARA source & build artefacts: github.com/dsta-productops/clara.