Nyx coding with Claude Code connector
WORKS WITH CLAUDE CODE

Claude Code, with a seat on your team.

The Clarity Forge connector gives every session your project's knowledge and your team's standards, turns the work into tracked tasks and artifacts, and answers when someone mentions it in a discussion.

01 · CONNECT

One command. Pick your agent.

Run cf-claude and choose a scoped agent, mapped to a team, a project and a folder on your machine. That choice decides everything that loads next.

$ cf-claude
▸ Connecting to Clarity Forge...
? Select agent
› Gideon · Engineering · /repos/web-app
Nova · Marketing · /content
Iris · Product · /specs
✓ Agent: Gideon · Engineering
✓ Workspace mapped: /repos/web-app
02 · BRIEFED

It shows up with a head start.

Before the first prompt, the connector syncs the agent's skills and your project's knowledge into the session. Everything is indexed and organised so Claude Code knows what's available and pulls what it needs — not dumped wholesale into every prompt.

✓ Agent: Gideon · Engineering
▸ Syncing skills
├─ code_review_checklist
├─ typescript_conventions
└─ testing_standards
▸ Pulling project knowledge
├─ architecture_decision: Atomic design, React 19
├─ coding_standards: TypeScript strict, Vitest
└─ brand_guide: Voice, tone, terminology
✓ Session briefed
03 · START A TASK

Pick a task. It arrives with the details attached.

Choose from the project's task queue or describe a new one. The connector pulls the description and everything linked to it — attached files, related notes, goals and dependent tasks — so the agent starts with the specifics, not just the project.

? Select task › Build pricing comparison table
▸ Loading task context
├─ description + acceptance criteria
├─ 2 attached files
├─ linked note: Q3 pricing rationale
└─ linked goal: Launch new pricing page
✓ Task context ready · starting session
04 · CAPTURE

Every session starts where the last one ended.

When the session ends, what it worked out is written back: decisions, patterns and lessons, filed as knowledge where the next person finds them. One note per domain, updated in place. The tenth session inherits the nine before it, human or agent.

✓ Session complete
▸ Saving results
├─ task updated with outcome
└─ 3 files written to project
▸ Updating knowledge base
├─ component: PricingComparisonTable
├─ pattern: Pricing tier data structure
└─ decision: Table layout over card grid
✓ Synced to Clarity Forge
05 · FOLLOW UP

Mention it like anyone else on the project.

Tag an agent in a Clarity Forge discussion. It picks the mention up on its next poll, reads the thread with the project\'s full context behind it and replies. Then it goes back to its task queue. No new interface to learn. The same discussions your team already uses.

Engineering, marketing, product and design agents can work the same project at the same time. Each keeps its own domain knowledge, so what your engineer figures out lands in architecture.md and what your marketer figures out lands in market-insights.md.

Discussion

Q3 Pricing Update

Sarah Chen
Sarah Chen2m ago
@Nyx Can you review the pricing table component and suggest improvements to make it more accessible? We need to meet WCAG AA standards.
Nyx
NyxAgent2m ago
I've reviewed the pricing table at src/components/PricingTable.tsx. Three changes would bring it to WCAG AA: add scope attributes to header cells, ensure 4.5:1 contrast on the feature text, and add keyboard navigation for the comparison toggles. I can implement these if you'd like.
SCENARIOS

What people run it for

Claude Code writes files. That means it handles any work that produces documents, analysis or creative output — not just code.

Software engineering

Assign development tasks from Clarity Forge. Claude Code receives your architecture decisions, coding standards and dependency maps before writing a line of code. Ship the results and the learnings.

Content & marketing

Pass your brand guidelines, tone of voice and messaging framework to Claude. It drafts blog posts, white papers and competitive analyses that sound like you — and captures what resonates for next time.

Product & strategy

Feed Claude your user research, feature specs and design rationale. It produces requirements docs, acceptance criteria and competitive analysis grounded in what your team already knows.

Operations & process

Hand Claude your runbooks and process documentation. It produces SOPs, incident reports and capacity plans that build on your operational knowledge base over time.

TWO WAYS TO WORK

Autonomous or supervised. Your call.

The connector adapts to how you want to work, from fire-and-forget to hands-on-keyboard.

Autonomous

Assign and walk away

Create a task in Clarity Forge, assign it to an agent and let it run. Come back to finished work, captured knowledge and a summary of decisions made.

Agent polls for tasks automatically
Full project context injected before work begins
If a sync fails partway, state is saved and retried on the next run
Supervised

Work together, keep everything

Use Claude Code the way you normally would — but with your project context pre-loaded and every decision automatically captured back to Clarity Forge.

Project knowledge loaded into your session
cf-claude --resume picks up your last task with plan progress and context intact
Learnings captured when you close the session

Give Claude a project to belong to.

Install the connector and your next session starts with everything the last one learned.

Get started →