The Operator Workspace that runs every Stride engagement
The vault and agent stack the studio runs on — the same substrate installed for clients — built to prove the model on ourselves first.
Running a studio with AI agents in the loop surfaced the same problem clients have: context scattered across chats, docs, and memory, with every agent starting cold.
We needed a substrate where humans and agents share one source of truth — and we needed to run it ourselves before we'd sell it to anyone.
We built an Operator Workspace: a structured Obsidian vault holding SOPs, decisions, and project context, paired with an agent stack of twelve specialized agents (Vitrine, Vaulture, Atlas, Steward, and others).
Each agent has a defined job — research, drafting, review, operations — and all of them read from and write to the same vault, so context compounds instead of evaporating.
This very site was specced, reviewed, and hardened against that workspace; the agents on the page reviewed the page.
┌────────────── Operator Vault (~2,000 notes) ──────────────┐
│ SOPs · decisions · project context │
└───────────────┬───────────────────────────┬───────────────┘
read/write │ │ read/write
┌────────────────┴───────┐ ┌─────────┴───────────────┐
│ 12 specialized agents │ │ Stride team │
│ Vitrine · Vaulture · … │ │ (humans) │
└─────────────────────────┘ └──────────────────────────┘A single source of truth with roughly 2,000 notes that both the team and the agents operate against.
Twelve specialized agents running real work against the vault, not a demo.
The exact substrate we install for Operator Workspace clients — customized per client, same foundation.
- ▸We sell what we use — running the substrate ourselves is the credibility, not a case study we commissioned.
- ▸The vault is the contract: when humans and agents share one source of truth, every session gets smarter.
- ▸A workspace tuned to one team doesn't transplant raw — the substrate is the same, the customization is the work.