Changelog

#025: Custom Roles (RBAC), Knowledge Base & a Smarter Model Picker

Invent #025: Custom Roles with full RBAC and per-resource permissions, Knowledge Base with include and exclude URL scoping, and a model picker with the full spec card for every model.

Jun 26, 2026

#025: Custom Roles (RBAC), Knowledge Base & a Smarter Model Picker
Blog/Changelog/#025: Custom Roles (RBAC), Knowledge Base & a Smarter Model Picker

June 26, 2026

It's Friday, and that means new updates from Invent. This week's batch is all about control: who on your team can do what, what your Assistant knows, and which model is running the show.

We love hearing from business owners and agencies about how they keep scaling their businesses with Invent, and this week's updates put more of that control in your hands: custom roles for your team, a sharper Knowledge Base, and a model picker that shows you exactly what each model can do.

The Highlights

Custom Roles for your team (full RBAC, on every plan)

Most platforms call this RBAC (Role-Based Access Control) and lock it behind enterprise tiers. In Invent it ships to every workspace, with the flexibility to build any role your business actually needs on top of the four built-in ones (Admin, Developer, Manager, Agent). A "Business Support" role, a "QA Reviewer" role, a "Client Reviewer" role for an agency client, anything.

A composite view of Custom Roles in Invent: the Create role modal on the left with fields for Name (Business Support), Color, Description, and per-resource permission toggles (None / View / Manage), and the member assignment modal on the right showing built-in and custom roles assigned to jane@zydeer.com with a note that Manager, Agent, and custom roles get a white-labeled, client-facing experience.

Create a custom role with the exact permissions your team needs, then assign it to teammates like any built-in role.

What full RBAC unlocks for your business:

  • Flexibility: build any role your business needs (not the four someone else imagined) and mix custom roles with built-ins on the same team.
  • Security: lock down sensitive resources (billing, members, integrations, contacts) with per-resource None / View / Manage permissions, so people only see what they should.
  • White-labeled by default: Manager, Agent, and custom roles get a client-facing experience branded with your org, and never see Admin or Developer users.

Stop bending your team into someone else's role structure. Build the access model your business runs on.

→ Full setup walkthrough in the Custom Roles docs.

Your Knowledge Base, easier to overview at the page level

URL include and exclude paths have been in Invent for a while. What's new this week is the per-page visibility and control. See every URL your Assistant crawled, preview what got pulled in, and exclude or include any single page with one tap.

A composite view of the Invent Knowledge Base modal for docs.useinvent.com showing the Pages tab on the left with each crawled URL and per-page actions (Re-crawl, Exclude, Preview, Delete), and the Settings tab on the right with Mode set to Crawl Pages, Include Paths set to /blog, and Exclude Paths set to /pricing.

Manage every crawled page from the Pages tab, and keep your URL scope rules on the Settings tab.

What's new:

  • Per-page preview: see exactly what your Assistant learned from each URL, so the index never holds surprises.
  • One-tap include/exclude: add or remove individual pages from the index without re-scoping the entire crawl.
  • Status at a glance: every crawled page shows when it was last refreshed and how big it is.

Cleaner index, more accurate answers, no more wondering where the Assistant pulled that line from.

See what each model can do, before you pick

Pick a model in any chat and you now see exactly what it can do, so you can match it to the job. Hover any model to see its full spec card. If it can't handle the features you've enabled (auto-resolve, follow-ups, anything else), you'll get a warning before you commit.

A composite view of the Invent model picker on the personal chat showing the dropdown (Auto, ChatGPT, Claude, Gemini, Grok, More models) on the right and the Gemini 3 Flash hovercard on the left with Context (1M words), Speed, Intelligence, Supports (Text, Images, Documents, Audio, Video), and Released date.

Hover any model in the picker to see its full spec card: context window, speed, intelligence rating, supported inputs, and release date.

The card shows:

  • Context window and speed.
  • Intelligence rating (so you know how much horsepower you're spending).
  • Supported inputs: text, images, documents, audio, video.
  • Release date and knowledge cutoff.

Leave the picker on Auto to let your Assistant route to the best model, or pick by hand when you know what the task needs.

Why it matters

Most AI tools give you the AI and trust you'll figure out the rest. This week is about the opposite: giving you the dials so the AI runs the way you would run it. Custom Roles (full RBAC, on every plan) means your team works the way it already works, with the security to lock down sensitive resources and the flexibility to build any role your business needs. Knowledge Base scoping means the Assistant learns only from the pages you choose, so its answers stay accurate and on-topic. And the model picker spec card means you stop guessing about the model behind every chat. The AI runs the work. You set the rules.

Weekend reading

Three fresh reads from this week:

New to Invent? Build your first AI Assistant in minutes with our docs and step-by-step guides.

With love,
Invent Team

Conversational AI for Business Growth

Start Building Your Assistant For Free

No credit card required.

Keep reading

AI for Agencies: The Complete Guide to Reselling AI
Industry

AI for Agencies: The Complete Guide to Reselling AI

A complete guide to AI for agencies: resell white-label AI assistants, manage multiple clients, and grow recurring margin with Invent, with no per-seat fees.

Alix Gallardo
Alix Gallardo
Jun 26, 26
SMS for Business: Why It Still Works, and How to Use It
Industry

SMS for Business: Why It Still Works, and How to Use It

SMS still gets read like nothing else. See why texting works for business, where it dominates, the top use cases from OTP to marketing, and how to send SMS campaigns with Invent.

Alix Gallardo
Alix Gallardo
Jun 25, 26
Instagram AI Assistant: Automate Your DMs and Win More Customers
Product

Instagram AI Assistant: Automate Your DMs and Win More Customers

An Instagram AI assistant answers your DMs automatically, in your brand voice, day and night. See what it does, why it works, and how to set one up with Invent, no code.

Alix Gallardo
Alix Gallardo
Jun 23, 26
Invent vs Botmaker: The No-Code, SMB-Friendly Alternative (2026)
Industry

Invent vs Botmaker: The No-Code, SMB-Friendly Alternative (2026)

Compare Invent and Botmaker on pricing, channels, and setup, and see why a no-code AI platform that is free to start fits SMBs better than a $149 monthly floor.

Alix Gallardo
Alix Gallardo
Jun 23, 26
Invent vs Freshdesk: The No-Code AI Alternative for Customer Service (2026)
Industry

Invent vs Freshdesk: The No-Code AI Alternative for Customer Service (2026)

Compare Invent and Freshdesk on pricing, AI, channels, and setup, and why no-code, usage-based AI fits growing teams better than per-seat helpdesk fees.

Alix Gallardo
Alix Gallardo
Jun 23, 26
#024: API Access for Every Plan & Read Receipts
Changelog

#024: API Access for Every Plan & Read Receipts

Invent #024: API access opens on every plan, including the free tier, plus read receipts that show delivered and seen across WhatsApp, Instagram, and Messenger.

Jorge Trujillo
Jorge Trujillo
Jun 19, 26