Use cases
Workstate gives every team the same foundation. It connects what your team already keeps, such as code, documents, tickets and files. It also keeps a ledger: an append-only record of what the team decided, what broke and what it tried.
Your people keep the agents they already use: Claude Code, Claude, Cursor, Codex, any other MCP client, or agents you built. Those agents search the same memory and write to the same record. Workstate isn't an agent itself. It's the layer they share.
One foundation, every team
The habit is the same in every team:
- Search first. Before acting, the agent searches your sources and the ledger for what's known and what was decided.
- Cite. Every answer points to where it came from: a file and line range, a page, an issue or a ledger topic.
- Record after. When the work ends, the agent records the decision or finding as a ledger topic. When a decision changes, a new topic supersedes the old one, and the history stays intact.
What changes from team to team is what you connect and what you record.
| Team | Connect today | Record in the ledger |
|---|---|---|
| Engineering | GitHub, Confluence, Jira | Design decisions, incidents, defects, dead ends |
| IT & operations | Confluence, Jira, uploaded runbooks | Incidents, root causes, fixes that worked |
| Customer support | Jira, Confluence help pages, uploaded guides | Known issues, workarounds, support policies |
| Product | Confluence specs, Jira, uploaded research | Product decisions and the options you rejected |
| Legal & compliance | Uploaded policies and contracts | Interpretations and positions |
| People & HR | Uploaded handbooks and policies | Policy interpretations |
Slack Coming soon and Google Drive Coming soon are on the way. See Coming soon for the full list.
By team
EngineeringCodebase answers with citations, decision records, incident follow-through and onboarding.IT & operationsRunbooks for every on-call agent, and a record of what failed before and what fixed it.Customer supportKnown issues from Jira, help pages from Confluence and the incident record, cited in every answer.ProductSpecs, tickets and the history of every product decision, including why.Legal & compliancePolicies and contracts in their own namespace, with interpretations recorded as decisions.People & HRHandbook answers that stay consistent, with sensitive material kept apart.
By pattern
Patterns cut across teams. They describe how people and agents work together through the ledger.
Multi-agent coordinationSeveral agents on one system: search before acting, record after, supersede instead of rewriting.Human–agent handoffsHand work over through a ledger topic and its current state, not a chat transcript.One namespace per clientKeep each client or product apart, and pick the namespace with every request.
How to pick your first use case
Start with one team and three to five jobs. Good first jobs have three things in common:
- The sources already exist in GitHub, Confluence, Jira or files you can upload.
- The question comes up every week, so the time saved adds up.
- A wrong answer is easy to spot, because the answer carries a citation you can open.
Designing use cases has a scoring table and a one-hour workshop plan. The rollout guide covers the rest of a pilot.
Keep sensitive material apart
Everyone with access to a namespace sees everything indexed in it. Workstate has no per-document permissions, so put HR, legal and other restricted material in a namespace of its own. See Permissions.