type.com is an AI workspace where teams and AI agents work together in shared threads, connected to the tools you already use, like Slack, Gmail, Google Docs and your analytics. We started using the type.com AI workspace on August 23, 2026, and it now runs most of our client content work. This guide walks through how we set it up, the workflows we run, the numbers we could verify from our workspace, and how to copy the setup for your own agency.

Who we are

StartupCookie is a small, AI-native content and GTM agency for B2B SaaS startups. We do founder LinkedIn ghostwriting, thought leadership, product marketing, AEO (AI search) content, video and custom GTM agents. We cap the roster at six clients and programs start at $5,000 a month.

Barbara runs strategy, writing and every client relationship. Jelena Vojnović handles design production: blog visuals, data tables and video name tags. In type.com today we support 5 clients, each in its own space.

The problem before type.com

In 2025 our process was, in Barbara's words, "a ChatGPT chat box that I prompted for every single task." Every chat started from zero. Voice, banned phrases and client background had to be pasted in again, every time.

Early in 2026 we moved to Claude Code, with custom skills and a context folder for each client. Quality went up a lot. But everything lived on one laptop: the skills, the memory files, the client context. When Jelena needed brand context for a visual, someone had to re-explain it. Call transcripts sat in Granola, and email and Slack were separate from where the writing happened.

Why we chose type.com

Three things mattered in practice:

The 3 workflows we run every week

These are the three workflows that do the most for us. Today, every one starts by hand in a thread. We don't have them on a schedule yet.

1. Founder LinkedIn posts from real call transcripts

Connects: Granola (call transcripts), Gmail, Slack, and LinkedIn through the built-in browser.

  1. We pull the transcript of a founder call from Granola into the client's space.
  2. The agent loads that client's ghostwriter skill: voice notes, a ban list and a bank of the founder's own quotes.
  3. It drafts posts and checks every claim against the transcript and our sources.
  4. Barbara edits. Each correction is written back into the skill's feedback log, so the same mistake doesn't come back.
  5. The founder reviews, then we publish.
Diagram of the founder LinkedIn workflow on type.com: a Granola call transcript goes into the client's ghostwriter skill, the agent drafts and fact-checks posts in a shared thread, Barbara's edits are saved back to the skill, and the founder reviews before publishing
Founder posts: transcript in, edits saved back to the client's skill.

Result: one founder call produced 17 post drafts. Every correction goes into that client's feedback log, which is why the drafts keep getting closer to the founder's voice.

2. Blog and AEO content, with design in the same thread

Connects: Google Analytics, and Google Search Console through the browser.

  1. We start from data: a Google Analytics baseline and an audit of 30 buyer questions in AI search.
  2. That becomes a brief and an editorial plan.
  3. A blog-writer skill drafts the post. A lint script blocks phrasing we've banned, and a separate reviewer agent reads it cold.
  4. A fact-check pass checks every number and claim against the source.
  5. Jelena makes the hero image and data tables in the same thread, using a visuals skill that already knows the brand.
  6. The agent writes the CMS fields: title, meta description and slug.
Diagram of the blog and AEO content workflow on type.com: analytics and AI-search audit, brief, draft with lint script and reviewer agent, fact-check, designer visuals in the same thread, then CMS-ready fields
Blog and AEO content: research, draft, checks and visuals in one thread.

Result: one fact-check pass on 4 drafts caught and fixed 45 errors before the client saw them. On October 22 that client moved to a plan of 4 blog posts a week.

3. Client review and updates

Connects: type.com apps and docs.

  1. We drop a source document, like a product one-pager, into the client's space.
  2. The agent writes posts for each channel and makes the visuals.
  3. It builds a small review app: a weekly calendar and a list view, with a copy button on every caption.
  4. The client reviews it, and their feedback goes back into the same thread.
  5. We publish the approved posts.
Diagram of the client review workflow on type.com: one source document becomes channel-specific posts and visuals, which go into a review calendar app with copy-ready captions, then client feedback and publishing
Client review: one source doc, one review app, feedback in the same thread.

Result: one client one-pager became 9 posts across X, the brand's LinkedIn page and a spokesperson's LinkedIn, plus visuals for each one. On September 17 we saved that calendar format as a skill, and we've reused it for other clients since.

Results

Every number below comes from our type.com workspace as of October 27, 2026, counting only Barbara's and Jelena's activity. We haven't tracked hours saved or turnaround, so we don't claim them here.

MetricNumber
Clients supported in type.com5, each in its own space
Team messages per week1 (week of Aug 22) → 240 (week of Oct 17)
Threads we started per month1 in August, 8 in September, 32 in October
Active scheduled automations0. Everything starts by hand in a thread.
Post drafts from one founder call17
Errors caught in one 4-post fact-check45
Invoices reconciled against bank and card records69, in one session of 2 h 16 min

One thing we won't claim: our agency's growth over the past year started before type.com, with Claude Code. type.com is where that work became shareable across clients and between the two of us.

What we'd improve

How to set this up for your own agency

This is the order we'd follow if we started again.

  1. Create one space per client. Put that client's brand docs, voice notes and past work in it before you ask for anything.
  2. Bring your existing prompts and skills. If you've built Claude Code skills or saved prompts, upload them into the client's space instead of rewriting them.
  3. Connect the tools the work already touches. For us that's Gmail, Slack, Google Analytics and Granola. Connect only what the first workflow needs.
  4. Build one workflow end to end. Start with whatever you produce most often. For us that was founder LinkedIn posts.
  5. Save every correction back into the skill. Ask the agent to keep a log of your edits and a ban list, so the next draft starts from them.
  6. Turn your review format into a skill. Once a client review format works, save it so every client gets the same one.

Who type.com is a good fit for, and who it isn't

A good fit if you're a small agency or content team handling several clients or brands with different voices, and you want AI agents that work from shared context instead of a blank chat. It's especially useful if you already have prompts or Claude Code skills you want your whole team to use, and your work lives in Gmail, Slack and your analytics.

Not a fit if you want finished content with no editing, or you only need help with one document at a time. It also takes setup: skills, ban lists and review formats get better because you keep correcting them. And if a client contract limits AI tools, check that before you move their work in.

Frequently asked questions

What is type.com?

type.com is an AI workspace where teams and AI agents work together in shared threads. It connects to tools like Slack, Gmail, Google Docs and your analytics. Skills, documents and context are shared inside a space, so everyone on the team works from the same setup.

How do content marketing agencies use type.com?

We give each client their own space with its own skills, documents and connected tools. AI agents research, draft, fact-check and build review apps inside client threads, and the team edits and approves there. Our biggest uses are founder LinkedIn posts, blog and AEO content, and client review calendars.

What tools does type.com integrate with?

We use it with Gmail, Slack, Google Analytics, Granola and GitHub, and the agent can work in a browser for sites like LinkedIn and Search Console. type.com also connects to Google Docs and other apps, so check its integrations list for your own stack.

Is type.com the same as type.ai?

No. type.ai is a separate AI writing editor. type.com is an AI workspace for teams and agents.

How long did it take to set up?

We opened our first client space on August 23, 2026 and had four client spaces running by September 23. Moving a client's Claude Code skills over took one thread: we uploaded the zipped folders and the agent turned them into space skills.