CASi

CASi is Eclipse's AI engine. It powers every surface in the platform: Cortex Canvas, the content editor, Reddit, social media, and Content Profiles.

What separates CASi from a standard AI chat is accumulated context. CASi reads your brand description, workspace documents, competitor intelligence, research nodes, training material, prior articles, and previous CASi conversations for the active client. Your editorial standards compound over time rather than resetting with each session.

What CASi knows

CASi draws from four layers of context when it writes:

Your voice. CASi absorbs your brand context during onboarding and sharpens it with every correction you make. Ban a phrase on Monday, and it's gone from every draft that follows. Fix a paragraph structure in the editor, and queued drafts reflect the change immediately. You don't re-teach. CASi remembers.

Your gaps. CASi maps your published content against competitor coverage and identifies where you're missing. Each client profile surfaces an average of eight competitive gaps. Your next article starts with an angle that's documented, not guessed.

Your competitors. Eclipse tracks competitor brands across AI search engines, content, and positioning. CASi has access to this intelligence when writing, so it can position your content against specific rivals rather than writing in a vacuum.

Your corrections. Every edit you make is an instruction that persists. Flag a phrase in the editor and it's removed from every draft in the queue, across canvas, social, Reddit, and Content Profiles. One action, every surface.

Your workspace history. CASi can look across the active client account to find existing content, prior decisions, saved material, and previous CASi conversations. It can report on what is already in the workspace instead of relying on you to re-explain it.

CASi vs standard chat

On Cortex Canvas, you have access to both standard chat and CASi.

Standard chat is the focused assistant. Use it when you want full model choice and a narrower conversation around the material in front of you.

CASi is the strategic assistant. It uses the current canvas, but it can also draw on the wider client workspace. Brand terms, client notes, prior articles, research nodes, source material, and previous CASi conversations all stay available within the same thread. When CASi writes a sentence, it can point back to the material behind it.

Where CASi works

CASi operates across five surfaces with shared memory:

Cortex Canvas. Research, draft, and generate from connected nodes. CASi writes from everything on your canvas plus the wider client workspace and external sources the platform adds automatically.

Content editor. Highlight any section and instruct CASi to rewrite it. The edit draws from the same context as the original draft. The rest of the article stays untouched.

Reddit. CASi drafts replies grounded in your brand context, published articles, and competitor intelligence. Replies arrive in the community's tone, not corporate language. See Reddit.

Social media. CASi converts articles into posts fitted to each platform. Your phrase bans and tone rules carry across intact. See Social Content.

Content Profiles. Set your audience and rules once. CASi runs the pipeline on repeat, each batch sharper than the last because your corrections compound. See Content Profiles.

How CASi learns

CASi improves through use. Every interaction teaches it more about your brand.

• Phrase bans propagate across all surfaces. Ban a term in the editor and it disappears from canvas generation, social posts, Reddit replies, and Content Profile output.

• Inline edits carry forward. Restructure a paragraph or shift the tone, and CASi applies the same standard to the next article for that brand.

• 50+ controls governing tone, banned terms, citation format, and structure hold across sessions and team members. No decay between logins.

• Training material uploaded through Train CASi gives CASi access to your documentation, slide decks, investor decks, and internal knowledge. It draws from this when writing and verifying claims.

The result is a learning curve that works in your favour. Week one, you read every draft closely. By week four, you're scanning. The standards you set are already built into every piece of content CASi produces.

Memory across a client account

CASi memory is scoped to the active client.

That means CASi can remember previous CASi conversations across canvases for the same client, but it does not carry one client's history into another client's workspace.

Agency isolation

For agencies managing multiple clients, CASi maintains completely separate context per client workspace. Five brands means five distinct correction histories, five separate voice models, five isolated research bases. No bleed between accounts, even when the same team manages all of them.