Changelog
New features, improvements, and workflow updates from the team.
Release notes describe AI Agent when each update was published. For current setup and behavior, use the documentation.
An approval step is the one place a human authorises an agent to act, so the only safe default is to keep waiting. Approval nodes now proceed on an explicit approval and nothing else.
Onboarding could produce three tasks and send three near-identical emails a minute apart. Tasks ready for review are now batched into one email, and you pick whether that arrives immediately, daily, or weekly.
A connection can look connected and be completely unusable — expired credentials, a revoked token. Connections now carry a live health state, tell you which of the two it is, and let you test one on demand.
Turn business data into typed Tables for people and agents. Organize it with saved views, let an agent compute a cell or update a column in batches, and inspect how each result was produced.
Build Company Brain from your Tables, then ask questions across every Table in your workspace. Combine data across Tables, uncover trends, and get summaries in chat, then check when the data last synced and refresh it whenever you need a newer answer.
Chat now keeps the work behind an answer visible. Follow a live plan as it progresses, open cited knowledge and web sources inline, and return to the thread with its plan and web results intact.
Stop an in-progress workflow from the runs list or detail view. Its cancelled state updates immediately in either view.
Describe a change to a workflow in plain language and the builder makes it for you. New steps splice straight onto the canvas and the layout re-flows so the end stays last — no dragging required.
The message box is now identical across dashboard chat, the inbox, and ticket conversations — same attachments, same @-mentions, same quick actions. Learn it once, use it everywhere.
Tickets can now be deleted outright. Removing one cleans up its comments, subtasks, attachments, and links in a single step, behind a confirmation so nothing goes by accident.
Connect Telegram and its messages land in the same inbox as your email. Each one gets an AI-drafted reply you approve before it sends — with reactions, edits, and attachments all handled natively.
Connect Gmail and your agent studies how you actually write, distilling it into a reusable voice profile that shapes every reply it drafts. Review it, refresh it, or clear it whenever you like.
Your business context, competitors, and sources now live in a single master–detail view. Pick anything on the left, edit it inline on the right — no more hopping between tabs.
The workflow builder grew up. Drag from any block to add the next one, insert live data with a searchable picker, and see exactly which branch a condition took on its last run.
The dashboard opens onto a configurable command center — pinned chat, a live to-do list pulled straight from your tickets, an inbox digest of what needs you, and suggested next moves. Resize and arrange it to fit how you work.
Approve a ticket and an agent takes it from there. Each step is logged live as it runs, and any action that touches a connected service — posting, updating, publishing — pauses for your approval before proceeding.
All your run history is now in one place. Workflows, autopilots, and background system runs share the same master-detail panel — scroll the list, click a run, and drill into its steps or approve a pending action without leaving the page.
Your workspace now builds a structured company profile from your site — mission, vision, founding story, values, founders, headcount, and industry. Each field shows its source and confidence so you can see exactly where it came from and edit anything the crawl got wrong.
Drag a file into any chat and the agent reads it as part of the conversation. Attach a CSV, PDF, or text file and ask questions about it — the agent has the full contents in context, no need to upload it to your knowledge base first.
First-time setup is faster and more guided. A product tour walks you through every section of the dashboard with screenshots, and the use-case menu has redesigned previews that show what each automation actually looks like before you enable it.
Browse every available integration in one place and connect with a single click — no API keys to copy or rotate. Managed OAuth is live for Stripe, Notion, and Linear.
The Traces tab is now a full debugging surface. See each run by the prompt that started it, drill into individual steps, and get accurate cost and token counts broken down by model.
Organizations can now add members and assign roles — control who can view, edit, or administer agents, workflows, and workspace settings.
New workspaces open onto ready-to-run automations — enable competitor tracking, an SEO audit, or review mining in one click. On signup we crawl your whole site to pre-build the knowledge every agent draws from.
The agent builder gained advanced controls: turn on extended thinking for harder reasoning, set input and output guardrails to enforce your rules, and attach evaluators that score responses automatically.
A transparent credit system, end to end. Every agent action costs a defined number of credits, each plan includes a monthly pool, and a ledger tracks every debit and top-up — predictable pricing, not a per-token meter.
Connect Gmail and incoming mail flows straight into your agents. The agent reads the thread, drafts a reply, and queues it for your approval — nothing is sent without you.
Upload documents and connect data sources, and your agents retrieve and cite the most relevant content when answering or drafting proposals — answers grounded in your actual company data.
Agents can now propose board tickets straight from a conversation. An inline draft card appears for you to edit and accept — landing the ticket on your board without leaving the chat.
Attach platform-provided tool bundles to any agent with a toggle. The first two are Web Research — search the web and read any URL — and SEO Intelligence: keyword volume, rankings, SERP data, and backlinks.
Agents can work as a team. A supervisor delegates tasks to specialists — SEO, strategy, outreach research — and synthesizes their work into a single answer right in your chat.
Always-on automations that run on a schedule with no manual trigger. The flagship App Store Optimization autopilot monitors your listing and suggests metadata improvements on its own.
The agent detail view is now a resizable 3-pane console — conversation history, live chat, and an editable inspector — with a new Memory tab that shows what the agent knows about your business.
An approved ticket can trigger your agent to open a real GitHub pull request — the first live, reversible external action on the platform, always behind a human approval gate.
Drag-and-drop builders for creating agents and multi-step workflows. Model settings, instructions, and connected tools live on a canvas that maps directly to the live runtime — no separate deploy step.
AI Agent reads your website and extracts your colors, typography, personality, and tone of voice — so every agent starts with your company's context built in.
A real-time chat for talking to your agents. Responses stream as they're generated, threads persist across sessions, and you can inspect the agent's execution inline.
The foundation of the platform. Connect your tools, let AI surface weekly action proposals on a board, approve the ones you like, and watch approved actions run — every step human-gated.
The new platform arrives — a complete dashboard with agent management, a work board, an inbox, a skills library, and team settings. A clean-slate rebuild of everything that came before.
Visually build multi-step workflows by dragging and connecting agents, tools, and tasks. From simple sequences to complex logic, Flows let you orchestrate everything on one interactive canvas.
Combine multiple AI agents into coordinated units that pass information and execute complex, multi-step work together. Your AI just became collaborative.
From scraping LinkedIn and Reddit to generating images and analyzing TikTok profiles, your agents can now reach for 15+ built-in tools to automate research, content, and competitive intel.
A redesigned model picker built for clarity and speed. Compare model strengths, context limits, and performance tags like Fast, Reasoning, or Powerful — and pick what fits the task.
We upgraded how team executions are displayed. You'll now see input variables, agent roles, task flow, and real-time progress in one glanceable view.
Finding and assigning agents got easier. A unified search lets you browse agents, explore tasks, and assign whole teams from one interface — built for speed and flow.
Your agents can use real tools — web search, image generation, domain scraping — as part of their work. Assign the tools you want and the agent handles the rest with smart parameter mapping.
Build a fully custom team of AI agents to collaborate on complex tasks. Assign each a role — researcher, writer, strategist — and watch them work in sync. Like hiring a dream team, minus the payroll.
Attach documents to any agent and give it persistent knowledge it can reference across every task and conversation. Perfect for SOPs, product manuals, and onboarding guides — no need to re-upload.
Create SEO-optimized articles in your readers' native language. From Catalan to Korean, tailor your content strategy for global audiences.
Fine-tune your SEO titles, descriptions, and keywords. Whether you're optimizing or localizing, take full control over how your content ranks and resonates.
Let us build it for you. From discovery call to custom agent delivery, our white-glove service helps teams save 15+ hours per employee and cut costs by 30% in 90 days. You bring the workflow — we bring the AI.
Every plan now comes with built-in credits, and you can top up anytime with packs that never expire. Buy once, use forever — designed for predictability and scale.
Grok joins the lineup of models you can put to work, alongside OpenAI, Claude, and the rest.
Build your own custom AI agents, assign them specific tasks, and mention them right from the chat. Want an SEO writer? A research agent? An email composer? Just @mention and go.
OpenAI's latest GPT-4.1 models are in your toolkit, ready to power your agents.
Drop in any blog post and we'll rewrite it to match your brand's voice, tone, and SEO structure — rooted in your content strategy, not someone else's.
We don't normally call these out, but this one was big. We squashed bugs, polished rough edges, and sped everything up — from page loads to article generation.
Use the best model for the job — OpenAI, Claude, Gemini, or open-source options. Switch between them freely, or let the system pick the smartest one for your query.
Press ⌘K (or Ctrl+K) to open a global command menu from anywhere. Jump into chats, run actions, and change settings — lightning fast, no clicks needed.
Plug your Notion workspace directly into the chat. Import docs, pages, and databases, and ask anything about them like you would any other file.
Your agents can now search the internet in real time. Pull in the latest info, news, and facts right alongside your documents and chat.
Drop in multiple files — PDF, CSV, URLs — and ask questions across all of them at once. No switching tabs. One chat, all your context.
A cleaner way to start — pick general chat, an AI agent, or the SEO writer from one unified interface. Faster flows, fewer clicks.
New settings let you match your brand's tone of voice and image style effortlessly — and generate more ideas with one click, based on what you've already written.
Research, analyze, ideate, draft, and create articles from just your domain. Using real SEO data from DataForSEO and Ahrefs, it writes engaging, on-brand articles in a few clicks. Days of work, done in minutes.
Bring your entire website in to power proactive workflows — and we're actively exploring more.
Upload any document and chat with it directly.
All the latest OpenAI and DeepSeek models are now available to your agents.
A couple of free tools to help you work faster, no account required.
Our AI writer focused on high-ranking SEO content, now at 1.0.
The first draft of @writer — an AI writer built for high-ranking SEO content. Good SEO is evergreen, free organic traffic, but it's usually slow and hard to get right. We worked with SEO experts to make it simple.
The better the input, the better the output. Write your prompt and the AI asks follow-up questions to help make it sharper.
It's now easier to start a document chat — from the document page and from the create-document form.
A requested feature: turn enhanced search on or off per agent.
Have normal AI chats without leaving the app — no need to go elsewhere for one-off questions. (Thanks for the feature request.)
A big UI update that gives agents, document conversations, and uploaded documents their own space — and sets us up to build a more robust system.
Upload a CSV and ask questions across your data.
Responses now expose their source links, and they're included on PDF downloads too.
After testing, we've enabled document chat for free users too.
Use a URL as a data source and ask questions about it.
It's now easier to edit, reorder, and delete your agent's tasks.
Derive deep insights from your documents at light speed, staying in flow. Now you can talk with your PDFs.
You told us responses were getting cut off. We changed the algorithm so it can render 300% more — less should be cut off from here.
Share your agent conversations — people don't even need to log in to see them.
Export an agent conversation as PDF, image, or text.
Pick which model you want to use — GPT-3.5 or GPT-4 — plus a failsafe profile picture and other fixes.
Rename and archive conversations to keep your workspace clean and clear.
Let your agent run on its own — complete tasks continuously without having to be there.
We no longer auto-regenerate and reprioritize every task after each run, so your edits stick and the task list stays focused. The result: task execution is now 300% faster. You can add tasks manually or let AI generate them.
Conversations are now grouped by when they were created, and you can see which agents are currently running.
Edit the tasks an agent suggests before it runs them. (Highly requested — we listened.)