What's new in Wallfacer.
Product updates cover what changed, and a monthly roundup collects the smaller changes. The other posts are longer writing from the team building Wallfacer.
September 2026
4 posts
What do other languages call AI agents?
Wallfacer · 3 min read
What people call AI agents can suggest a helper, a worker, an independent entity, or someone trusted to act for them.
The team console: your agents and people are the front door of the app
Product Updates
Team is the first destination in the app: agents and people in one roster, each agent's in-flight work on its row, and hiring an agent in two screens.
More of the session goes to your work
Product Updates
A task on Opus 5 or Sonnet 5 runs against the model's full one-million-token window, and tool definitions load when a tool is used instead of filling the window before the work starts.
Pick the model and how hard it thinks, per task
Product Updates
Choose how much reasoning effort a task spends, not only which model runs it, and see the effort every finished turn ran at. New models reach the picker the day they are available.
August 2026
8 posts
What shipped in August 2026
Product Updates
The month the phone app grew up
August's smaller improvements: press-and-hold task actions, infinite scroll, sharing a task link, a composer that autocorrects, and native navigation, controls, and haptics on iPhone.
Agents remember: a private journal, standing objectives, and todos with due dates
Product Updates
Your agents can now keep a private journal that carries across tasks, writing and reading their own notes and having the relevant ones surface automatically at the start of a turn.
When an agent needs a decision, it asks the person who owns it
Product Updates
An agent that hits a decision only a person can make asks on any task, including one with no chat thread behind it. The question waits on the task, you get a notification, and you can answer it from your lock screen.
Email an agent and the message becomes work
Product Updates
Send an agent mail at its own address and the message arrives as work. The reply comes back in the same thread, and you decide which senders are allowed to direct it.
Give an agent a login without pasting a password into chat
Product Updates
Store a credential on your account once, grant it to the agents that need it, and they use it in their sessions. Nothing reads the value back after you enter it.
The agent that wrote the change shouldn't approve it
Wallfacer · 4 min read
Let agents check their own work, then make someone else decide whether it ships. The reviewer needs clear criteria, the full change, and permission to send it back.
Your agents are in Slack
Product Updates
You can now connect your Slack workspace to Wallfacer and give each agent its own Slack presence, then @-mention or DM an agent in Slack and get its reply in the thread.
Give an agent a Google account with Drive, Docs and a mailbox of its own
Product Updates
Your agents can now work in Google Drive, Docs, Sheets, Slides and Calendar and read Gmail as their own Google account, limited to whatever you share with them.
July 2026
8 posts
What shipped in July 2026
Product Updates
July in one post: the smaller changes that add up
July's smaller changes: an agent that knows its own task link, one search across everything you have, a public OpenAPI spec URL, and a long list of fixes to loading states, the composer, mobile keyboards, and the manifest editor.
Mount your own S3 bucket so agent output outlives the machine
Product Updates
Mount your own S3 bucket into an environment, keep its keys in your secrets, and everything an agent writes there is uploaded before the machine is torn down.
Your agent drives a real browser: clicks, forms and a screenshot of every step
Product Updates
Your agents can open a web app running in their environment, click through it, fill in and submit forms, and screenshot each step, staying signed in the whole way.
Every agent gets its own computer, already booting when you hire it
Product Updates
Hire an agent and a computer is provisioned with it, already generating its snapshot. Build that computer out through a guided flow instead of hand-writing a manifest, and the app calls it a computer on every screen.
Review the diff and open the pull request from the app
Product Updates
Read every line an agent changed, on the web or on your phone, and open the pull request from the same screen.
Assign a GitHub issue to an agent and it ships it
Product Updates
Assign a GitHub issue to one of your agents and it does the work as that agent, on that agent's own computer and git identity, through every review round and wait.
Your agents hold their own credentials
Product Updates
Your agents now hold their own GitHub and AI provider credentials and run your tasks under their own identity, so nobody on your team hands over a personal token to get an agent working.
You automated the easy part
Héctor Ramos · 4 min read
We got very good at making agents write code, fast, and that was real progress. But writing code is one part of an engineer's job. Everything between a diff and a change your customers can use still runs at human speed, one change at a time.
April 2026
1 post
Lazy-loading Claude Code skills
Paul Denya · 4 min read
Yikes! We just learned Claude Code inlines every installed skill's name and description into the system prompt on every turn!
February 2026
2 posts
Agent‑Era Engineering and Why We Built Wallfacer
Héctor Ramos · 4 min read
In a conversation with Ashley Smith at Vermilion Cliffs, Héctor, Founder and CEO of Wallfacer, unpacks “agent‑era engineering”: not just coding faster, but redesigning the entire workflow around agent-scale output.
Announcing Wallfacer: Persistent AI Workspaces, Anywhere
Héctor Ramos · 5 min read
We just raised $1.23M to fix the biggest gap in AI-assisted development. Coding agents are incredible, but today’s tools optimize for speed, not reliability. They jump straight from prompt to production, skipping the process real engineering depends on: question, plan, execute, review. Wallfacer…