The last blog update covered the plugin system and the mailbox module. A lot has shipped since then — here's what landed in main and is ready for testing.
Contacts: filtered list, sort, and ⌘K search
The contacts view got a full rebuild. Instead of a basic table — a viewport-fit list with a complete filter set, per-column sorting, and keyboard-first global search (⌘K).
Contact profiles are now fully editable: phone, notes, custom fields. Email stays locked after the first save — one contact, one address, no duplicates.
- Filters and sort on every column
- ⌘K — fast keyboard search
- Editable profile: phone, notes, custom fields
- Email locked after save (deduplication)
AI Compose + Pokelo RAG
This is the change that makes a difference in daily work. AI Compose previously only had the email thread as context — it knew what you wrote and what the client replied. Now it has access to the company knowledge base through Pokelo RAG.
In practice: when drafting a reply to an inquiry, AI can pull from product docs, customer service procedures, pricing — anything you've loaded into Pokelo. The suggested reply isn't generic anymore, it's grounded in your company's actual data.
The second part: first outbound drafts. When a new lead fills a form on a landing page, AI Compose automatically generates a suggested first message — factoring in the form data and relevant context from the knowledge base.
- RAG from Pokelo — AI knows what's in the company's documents
- Suggestions are grounded, not generic
- Auto-draft of the first message after a new form submission
- BYOK — bring your own API key (OpenAI-compatible)
Listmonk: email campaigns inside the CRM
The Listmonk plugin moved from a basic integration to a full campaign module. The campaign editor is now directly inside the CRM interface — write content, generate an AI draft, preview, and send without leaving the app.
CRM forms can now be mapped directly to Listmonk lists. Lead fills the form → lands on the right mailing list → receives the sequence. No manual transfers.
- Campaign editor inside the CRM UI
- AI-generated campaign content
- Form-to-list mapping
- Preview before sending
Pipeline: badge on lead cards
Small change, big impact. Lead cards in the Kanban now show a badge when there's an unanswered email from the client. You don't have to open every lead to check — it's visible at the pipeline level.
MCP: new tools for AI agents
The built-in MCP server got two new tools: reading email threads and sending mail attributed to a specific lead. An AI agent can now not only read CRM data but actively participate in correspondence — within permission boundaries, with a full audit trail.
What's next?
MVP is in the final polish stage. We're collecting beta testers — if you run B2B sales and are looking for a CRM that doesn't generate more work than it saves, reach out. The first group gets early access and direct input on the product direction.
Want to fix one concrete process?
We start with a conversation about data, context, and where AI can actually help — not another chat window beside the company.
We reply within one business day · office@bearly.pro
More from the blog
Why We Stopped Using AI as a Tool and Started Building an Environment for It
For a long time, the human was the integration layer between AI and the company. How we changed our approach and started building an environment for agents.
Why Your AI Sales Agent Isn’t Working — The Model Isn’t the Problem, Your Data Is
AI sales agents fail when CRM data is fragmented, duplicated, and cut off from external signals — not because the model is weak. How to fix the data layer first.
Magic prompts are bullshit — what actually changes how you work with AI
The obsession with prompts is a symptom of misunderstanding how AI models work. The real difference is the system: RAG, skills, memory, per-project instructions. The prompt is the last 20%.
