Jessie
The Asana-facing team agent for consultant work, reporting, monitoring, and review-ready outputs.
Internal system guide
Agency OS Architecture.
See how you and Jessie work together: how Asana, live data, client context, Cloudflare, and Platform Build fit around the work you do.
Choose Your Path
Two ways in, depending on what you're here for.
Using Jessie →
You brief Jessie through Asana and use her output. Start here for how to set up a task, what each skill does, and how to read and trust what comes back.
Building Agency OS →
You maintain or extend the system. Start here for the architecture, data plane, decisions, and how to add a skill.
What This Is
Agency OS
The platform around Jessie: context, source assets, data retrieval, run state, artifacts, observability, and architecture records.
Platform Build
The separate improvement loop where Agency OS development tasks are logged, built, reviewed, and merged.
This is internal documentation, written for you. It starts with what Jessie and the platform actually do today, then points to the technical source references if you need to extend or audit the system.
Who Does What
| Actor or layer | Plain-language role | Current boundary |
|---|---|---|
| Consultant | Creates or assigns work in Asana and adds task-specific notes. | Approves client delivery and any real-world change. |
| Jessie | Reads the Asana task, selects a skill, gathers context/evidence, calls the LLM through AI Gateway, and posts review output. | Does not publish to clients or change external systems today. |
| Cloudflare Worker/control plane | Receives Asana events, queues work, stores run state, loads bindings, and calls tools. | Runtime system of record for run state, not the human-facing work board. |
| AI Gateway | Routes approved model calls and records model-call observability metadata. | Not an actor. It is the model-call route and observability layer. |
| Codex/build agents | Work on Agency OS platform development tasks from the Platform Build loop. | May build in branches/worktrees. Merge requires human approval. |
Current Boundaries
Implemented or gated capabilities
- Run eligible reporting and analysis code only within its governed shadow or pilot cohort; each skill badge shows the current release state.
- Retrieve approved client context and task-specific consultant notes; semantic recommendation influence remains gated until its evidence threshold is met.
- Retrieve configured read-only data from GA4, Search Console, Google Ads, Meta, Browser Run, R2, and D1 where available.
- Route enabled insight-synthesis paths through Cloudflare AI Gateway; the broader production model rollout remains blocked by its recorded safety prerequisites.
- Prepare consultant-review Asana comments and internal artifacts for an approved cohort; this is not direct client delivery.
Jessie cannot do today
- Publish directly to clients.
- Change websites, tracking, ad accounts, budgets, campaigns, or dashboards.
- Treat consultant notes as permanent approved client context.
- Create or store secrets.
- Merge Platform Build code changes without approval.
Recommended Reading Path
1. Use Jessie
Start with task examples and prompt templates.
2. Understand Context
Learn how approved context differs from task notes.
3. Follow The Flow
See the Asana-to-report runtime path.
4. Pick A Skill
8 skill contract pages explain routing, rollout status, and outputs.
5. See The Destination
Review the working product, governance, and resilience vision.
6. Improve The Platform
Use Platform Build for Agency OS development work.