Changelog
Follow up on the latest improvements and updates.
RSS
New
DeepSeek V4 Flash 0731
is now in the model selector for chat, apps, agents, and workflows. - The model's 1-million-token context window makes it a fast, long-context model built for coding and multi-step tool use — a good pick when there's lots of material to work through.
Self-Reporting Credit Usage
Ask about your own credit usage right in chat and get an answer immediately — no admin access required.
Phone Agent: Pause, Duplicate, Reassign
Three new admin capabilities for phone agents:
- Pauseputs an agent on do-not-disturb. It stops answering calls but keeps its number and its call history, and you can bring it back anytime.
- Duplicatemeans you can copy a configured agent onto a new number. Build one front-desk attendant, then replicate it across as many numbers as a client needs instead of setting each one up from scratch.
- Reassignmoves an agent to a different tenant under the same MSP, keeping its number.
Improved
Present an Artifact Fullscreen
- Artifacts have a Present control in the viewer header. It puts the artifact into full browser fullscreen and hides the header and chat pane, so what's left on screen is just the artifact. Press Escape to come back.
Auto-Tool Tools Function Transparency
- When Auto-Tool is on, Hatz chooses the tools for you — switch Auto-Tool off in the Tools menu to choose them yourself. A notice at the top: "Turn off Auto-Tool to select tools" now appears in both the chat tools picker and the Workshop agent editor as a reminder.
Hatz Activate is your AI-as-a-service rollout command center, built into the Hatz Admin dashboard. It's not a funnel you run once — it's a set of tools you keep coming back to for every customer: continuous activation. Scan for Shadow AI, refresh use cases, and regenerate resources as your clients adopt Hatz over time.
Who Has Access
Primary Admins, Admins, and Sales Users
- Sales User is a new role available under Admin > Settings > User and Roles
- Find Activate by going to admin.hatz.ai and clicking on Activate in the top nav bar
Bring In Your Customers
Start by getting every customer into one place. Sync from HaloPSA, ConnectWise, or Autotask, or upload a CSV — all customers land in one table with status, source, and tenant match on each row, so you can tell at a glance who's already a Hatz tenant and who isn't.
From there, run enrichment to add industry, company size, and data exposure risk, giving you a fuller picture before you reach out.
Size the Opportunity
A Revenue Calculator estimates your potential earnings based on customer count, margin, and package, so you can gauge what's on the table as you work each account.
Readiness & Shadow AI — Start Conversations With Authority
A customer's readiness and risk don't stay fixed, so neither should your read on them. The 10-question AI readiness assessment isn't one-and-done. Retake it any time, and every attempt is saved, so you can see how readiness has shifted since the last conversation.
Shadow AI Scan works the same way: it shows how many devices are using free consumer AI tools, which tools, and how risky, and you can re-run it whenever circumstances change rather than relying on a single snapshot to stay ahead of Shadow AI creep over time after Hatz is deployed to the client.
Use Cases and Resources — Built for This Customer, Updated as You Learn More
Nothing here is static. Use Cases proposes AI opportunities targeted to each customer's business, and you can add context or regenerate them any time as you learn more about what they actually need.
Custom Resources turns that into something you can hand a customer, or use internally — a sales deck, one-pager, objection guide, or onboarding template, personalized with their info, your branding, and any "Context for AI" notes you've added.
If a draft isn't quite right, tell the AI what to change — the whole document or just one section — and it revises on the spot instead of making you start over.
Reach Out
Draft and send emails straight from Hatz, cc'd to yourself, with any custom resources attached — and each one is saved to the customer's record, so the outreach history stays with them instead of living in your inbox.
Helpful Links
- This Activate User Guide will help you launch AI-as-a-service in your MSP
- Docs Article about Hatz Activate
- You can read more on our website, and here's more written by our team.
July 30, 2026
New
AutoTool v2
An updated version of the Auto-Tool feature makes automatic tool use more reliable as your conversation evolves through:
- Better tool selection based on each request
- New tools can be discovered as the conversation evolves
- Improved recovery when tools change or fail
Auto-Tool Default
- The Auto-Tool toggle in chat now also defaults to on for new chats, so you don't have to switch it on yourself every time. It's moved into the tools dropdown, right where you'd expect it.
Phone Agent: New Usage-Based Model
Starting on August 1st, the Hatz Phone Agent no longer requires a license to get started. Create a phone number and you're billed for what you use.
- Admins can now set a monthly minute limit per number to keep usage predictable, with a simple usage bar showing minutes used against that limit.
- Call logs now show minutes used this month right alongside transcripts and call counts, so it's clear what's being billed.
- Post-call summary emails now send strictly after the call ends, instead of during in-call handling, for more consistent behavior.
Improved
- MSP Workshop categories now include agents. MSP admins browsing agents can now see MSP-only categorized items, matching how it already worked for other Workshop item types.
Fixed
- Custom MCP integrations that need extra permission mid-conversation now prompt you correctly, instead of failing silently.
- Fixed an infrequent issue where workflows with optional inputs left blank would fail to run — they now complete successfully.
New
2 New LLMs
Kimi K3
, Moonshot AI's most capable model is now available in the model selector for chats, agents, and workflows. Its responses take a little longer to arrive, but extra thinking behind them delivers stronger, more thorough output. "
Kimi K3 is a 2.8T-parameter model built on our Kimi Delta Attention and Attention Residuals, with native vision capabilities and a 1-million-token context window. It is the world's first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning."
– Moonshot AI
GLM 5.2 Fast
(ZAI) is now available in the model selector for chats, apps, agents, and workflows. A speed-optimized serving tier of GLM-5.2, hosted by Baseten and built for real-time, latency-sensitive agent workloads. "
GLM-5.2 Fast: our GLM-5.2 Model API designed for the most demanding real-time use cases. The Fast tier delivers 2-3x higher TPS than the standard GLM-5.2 MAPI.
" – Baseten
New
Claude Opus 5 is now available in the model selector for chat, apps, agents, and workflows.
Photo provided by Anthropic
"Claude Opus 5 provides greatly improved performance for the same cost as its predecessor, Opus 4.8. .. Opus 5 excels on valuable software engineering tasks. For example, on Frontier-Bench v0.1, Opus 5 surpasses all other models, and more than doubles Opus 4.8’s performance at a lower cost per task."
. -AnthropicNew
New LLMs
- Gemini 3.6 Flashis now available in the model selector for chat, apps, agents, and workflows
"Our workhorse model that delivers better coding, knowledge work, and multimodal performance. According to the Artificial Analysis Index, it reduces output token usage by 17% compared to 3.5 Flash, and in some benchmarks like DeepSWE by Datacurve, we observe up to 65%, all at a lower cost per output token."
— Google
- Gemini 3.5Flash-Liteis also available in the model catalog. A lower-cost option in the Gemini family for high-volume, everyday tasks — with vision, tools, and thinking.
"
Our fastest, most cost-effective 3.5-class model, delivering 350 output tokens per second according to the Artificial Analysis Index, also significantly outperforming prior Flash-Lite generations in agentic workflows.
" — Google
- Inkling(Thinking Machines Lab) is fully available in the model selector for chat, apps, agents, and workflows.
"Inkling is a Mixture-of-Experts transformer with 975B total parameters, 41B active. It supports a context window of up to 1M tokens. It was pre-trained on 45 trillion tokens of text, images, audio and video. It is the first in a family of models of different sizes: alongside it we are sharing a preview of Inkling-Small, a lighter-weight model with 12B active parameters, trained with a similar recipe, that achieves strong performance with even lower cost and latency."
— Thinking Machines
MSP & Admin Tools
- MSP Admins and client admins can now get usage insights directly through Hatz — who's using the most credits, daily and per-model usage trends, a breakdown by user/model/feature over any date range, and which users are building the most apps and workflows.
MSP Admins see this across every tenant they manage; client admins see it for their own organization.
MSP Admins can check credit usage across every tenant they manage in one place, with an early heads-up on anyone nearing their limit.
Renaming Artifacts
- Click "Rename" from any artifact you own, at any time, to give it a clearer title — it updates everywhere you'd expect (Workshop, your chat, shared views), while your content, versions, and public link stay exactly as they were.
120 New Hatz Official Workshop Items
- 120 new Hatz Official workshop items were added to the Discover tab, and 9 new categories are available for tagging: Universal, Research, Sales, Customer, Support, Nonprofits, Executive Leadership, IT, Project Management, Tax & Accounting
Hacker News Integration
- Re-introduced a faster, more reliable Hacker News integration in the tool base
Improved
- Preview any Word or PowerPoint document Hatz creates for you right in the chat — no need to download to see files you produced.
- Anyone you send a public link to sees a clean, full view of the document, and you can rotate or revoke access whenever you need to.
- Various updates to custom MCP reliability.
Phone Agent New Setup + Upgrades
- Redesigned create & edit experience Creating and editing a phone agent is now a streamlined, modal-first flow — cleaner advanced settings, an improved workflow picker, voice selection, and a review step before you save.
New Phone Agent Setup:
Existing Agent Edit:
- Post-call summary emails Enable summary emails on your agent to get a recap delivered after every call.
- Smarter call handling If a caller goes silent, the agent now politely re-engages before ending the call, and we fixed an issue where the first word of the agent's greeting could be clipped. Call logs now show times in your local timezone with clearer status badges.
- Caller Blocklist Phone agents can now block spam and unwanted callers. Manage blocked numbers from the new Blocklist Manager in your agent's settings, or block a number in one click straight from the call logs. Blocked callers are screened out before the AI engages, and their calls are automatically labeled with a new Blocked status in your logs.
Artifact Upgrades
- Public artifact links — share artifacts with anyone, no Hatz login required. Generate a public link and recipients get a secure, view-only page; rotate or revoke the link at any time.
Account Teams
- Account Teams let you limit which members of your organization can access which customer tenants. When Account Team enforcement is on, a team member who is not an admin can only open and manage the tenants they have been added to. This is useful when different people own different books of business and you want each person to see only their own accounts.
New LLM
- Grok 4.5 is now available in the model catalog
"SpaceXAI's smartest model built to excel at coding, agentic tasks, and knowledge work. It's our strongest model ever and was trained alongside Cursor. Grok 4.5 was trained on datasets spanning knowledge in coding, science, engineering, and math. With both intelligent and efficient reasoning, Grok 4.5 excels at real engineering tasks and exceeds comparable leading models at these tasks."
- SpaceXAIImprovements
- Workflows with expired or broken integration credentials are now clearly flagged reconnect required, and triggers can no longer dispatch through unauthorized connections.
- Items shared by a group editor are now visible to all group members.
- Fixed: workflow canvas appearing empty
- Custom platform roles now correctly see the Integrations & Tools page.
- New guardrails prevent publishing apps or workflows pointed at retired models, and existing apps and workflows referencing a retired model are automatically healed to its successor — nothing silently breaks when a model is sunset.
- File download links in chat output now work reliably, long-running code executions recover from sandbox timeouts, and deep research no longer hangs on long runs.
New
New LLMs
- Fable 5 (Anthropic) has been re-enabled
"Claude Fable 5 introduces our 5th model generation for your most ambitious work. Tackle days-long, complex, and asynchronous tasks previous models couldn't sustain." — Anthropic
Claude Fable 5 is Anthropic's most capable generally available model, built for long-horizon coding projects, deep research, and multi-stage enterprise workflows. It can plan across stages, delegate to sub-agents, and check its own work — including using vision to evaluate outputs against the original goal. It's state-of-the-art on coding, knowledge work, and computer use.
⚠️ Note: Claude Fable 5 consumes significantly more credits than other models — approximately 12x our baseline. It also requires special 30-day data retention for safety monitoring. Learn more about Fable's data retention practices.
- GPT-5.6 Sol (OpenA)
- GPT-5.6 Terra (OpenAI)
- GPT-5.6 Luna (OpenAI)
"GPT‑5.6 Sol is our strongest model yet. To give a preview of model performance, we share a set of evaluations highlighting improved agentic capabilities in coding, biology, and cybersecurity" - OpenAI
Artifacts
- An artifact is a reusable item created by AI and distributed through Workshop, group shares, or public links.
- Ask Hatz to build a report, slide deck, tracker, or interactive web page, and it's created as an artifact you can preview and refine right in the conversation.
- Creates Code, PDF, DOCX, PPTX, and HTML
Compliance & Logs
- Admins can review a complete, searchable record of every AI interaction in their organization.
- See each invocation — the user, their tenant, the model used, and the chat, app, or agent it came from — then filter by user or date range and open any entry to inspect its content.
- Records can be exported for audits and reporting.
- MSP Admins have full control over who has access to this feature and can choose whether or not to enable it per tenant.
Search Hatz Docs From Hatz Chat
- Hatz users can now reach Hatz's own knowledge and capabilities through MCP, and without enabling tools.
- Get real time answers based on searches of our up to date documentation.
Improved
Phone agent improvements:
- Phone agents now report the correct, local time to callers
- Natural language interruptions - callers can now speak over the agent and it intelligently classifies whether that speech is a real interruption, so conversations feel more natural.
New
- Phone Agents now recognize and greets returning callers by name, using context from prior calls. Only the caller name is retained between calls from the same phone number. Can be enabled or restricted by Admins.
- Admins can now set a maximum call length for their AI phone agents, giving them control over how long the AI stays on a call. When enabled, the agent gracefully wraps up the conversation with a friendly message before the time limit is reached, then ends the call. This feature is off by default and only affects the AI portion of a call — calls that have been transferred to a live person are never cut off.
- Admins can now set business hours for a phone agent (days, hours, and timezone). Outside of those hours, the agent switches to after-hours mode — it can still chat and take a message for follow-up, but it won't transfer the caller.
- You can now use one of your existing agents as the brain behind a phone agent instead of building a phone agent from scratch. If an agent is not eligible to be selected for the phone agent, the reason is displayed.
Improved
- When you don't manually choose a model (you're on "Auto-LLM") the system decides which underlying model handles your request based on what you're asking for. We swapped in newer, better-performing models behind that automatic selection for Lite and Performance modes, so you get better output quality automatically.
- If an agent hits a step it can't get past, it now stops itself and tells you specifically what it was trying to do and where it got stuck instead of quietly retrying in the background indefinitely.
New
- Claude Sonnet 5 is now available in the model selector for chat, apps, agents, and workflows.
Sonnet 5 narrows the gap: its performance is close to that of Opus 4.8, but at lower prices. It’s a substantial improvement over its predecessor, Sonnet 4.6, on important aspects of agentic performance like reasoning, tool use, coding, and knowledge work.
- Anthropic
Scores for Sonnet 5 on a variety of evaluations compared to those of Sonnet 4.6 and Opus 4.8 (a more generally capable model, for reference).
- Anthropic-Seven new LLMs added to the model selector. These models are US-based, lightning-fast, and built for efficiency — delivering powerful AI performance with low credit usage.
GLM 5.2
(Z AI) — Z AI's sharpest thinker yet, equally at home reasoning, coding, and running your tools.GLM 5.1
(Z AI) — A reliable all-rounder from Z AI, ready for whatever reasoning or coding task comes its way.Kimi K2.7 Code
(Moonshot AI) — Moonshot's coding specialist, built to write, debug, and ship without missing a beat.Kimi K2.6
(Moonshot AI) — A long-memory multitasker that reads, sees, and summarizes with ease.DeepSeek V4 Pro
(DeepSeek) — DeepSeek's flagship reasoner, diving deep into hard problems and surfacing with the right answer.Nemotron 3 Ultra
(NVIDIA) — NVIDIA's biggest brain, built to power through the toughest workloads.gpt-oss-120b
(OpenAI) — OpenAI's open-weight model, bringing their signature smarts to everyone.Improved
- The Microsoft 365 integration got several upgrades: browse and search calendar events across any date range (previously capped at about a week or 50 events), filter email search by date range, read/unread status, importance, and attachments, create and move Outlook mail folders, and read and search from shared mailboxes you have access to.
Load More
→