This release is about making agents harder to knock off course. Updated models in the GTM and marketing harnesses, tighter execution, and a round of UI and reliability fixes. Same agents, fewer edits, fewer ways for a run to go wrong, smoother day to day.
We launched Cursor GTM: a done-for-you / done-with-you service that builds a GTM engine on Cursor for companies in the ecosystem. Skills, agentic workflows, tool calling, and guardrails, configured with you. You own the repo. Strategy stays with your team.
Two production agents for paid media are live. Search Ads Sherpa (Google Ads) and Meta Ads Ace read the account, draft the work, and queue changes paused for your review. Spend does not move until you say so.
We launched Metaflow 2.0. Traditional workflows are out. The platform is fully agentic: one harness around GTM work, not a canvas of blocks you have to wire yourself. Faster to value, and integrated from day one.
Agents now run on shared company context. Knowledge, onboarding, memory, and library are native, not add-ons you stitch together.
Agents can now call Skills: reusable, inspectable operating procedures instead of a hidden prompt. Built-in marketing skills ship with the platform. Your team can add their own (playbooks, SOPs, and internal groupings) so every agent in the workspace works the same way your company actually works.
We have migrated our third-party integration service for agents to connect with other apps. So moving forward our focus is on depth, hence we now support a focused set of apps. Formerly we prioritized coverage over reliability which is a reason for our previous infra decision. Now given we now have clear insight and visibility on the handful of apps where we have seen the largest concentration of usage. Our goal is to focus on those, and continue to enhance and deepen integration in those apps.
Agents now offer all the latest models. Plus the agent architecture is upgraded with drastically better performance. The agent is now context-efficient with better summarization, and architecture now supports long running tasks even with smaller models.
Entity Optimization is now available as a tool your agent can run. This is. NLP based optimization technique that's usually gated as a priced feature in other products. We offer an advanced version of that without additional cost. This lets your content not only sound and feel legible but actually makes it semantically richer so that its LLM friendly and found to boost visibitily and citation by atleast 3Xing for team collaboration
You can now run tasks at scale using Growth Runs feature that gives you a matrix view. This is where you can map your triggers and layout agents and flows are columns that you can run at bulk.
Multi-tenancy is now part of all plans with role provisioning for team collaboration
Now, Agents can be added as part of a flow. More importantly unlocking Multi-Agent Workflows in Metaflow
Visualize any data into beautiful charts in Agents using the newly added visualizer tool
Any workflow can now be turned into a reusable tool. Agents orchestrate, reuse, and scale workflows as primitives.
GPT-5 is now embedded inside flows and agents, delivering deeper reasoning, stronger orchestration, and higher reliability.
Launched support for running your own custom MCP servers, extending integrations beyond the built-ins.
Metaflow now integrates with 2,500+ apps via the Model Context Protocol (MCP).
Flow now features a 'Portal' block that let's another Flow to be plugged within as a sub-workflow.
Agents can now save generated outputs into doc, bringing workflow continuity and paving a path for workspace-as-memory.
Brought live SEO data so our users and their agents can retrieve relevant search and keyword insights natively.
Launched support for running your own custom MCP servers, extending integrations beyond the built-ins.
Native Notion Integration for instantly publishing pSEO content.
Invite walls came down. Anyone can now access the platform.
Launched our closed beta with a fully functional flow editor, featuring 17 block types for multi-model automation, 25 frontier text, image and vision models.
Metaflow, Inc. was officially incorporated.
Early conversations with growth marketers and founders validated the need: no-code workflows powered by AI.
The thesis came from a simple observation: single LLM calls whisper fragments of intelligence. When these moments of micro-cognition are woven into continuity with more durable structure, they unlock the automation of complex, deeper knowledge work—less about productivity alone, and more about freeing humans from the desk toward more rewarding pursuits.