Oi
Oi is a shared operating layer for teams using AI assistants. It organizes connections, context, skills, guardrails and workflows so a team can reuse its setup across tools such as ChatGPT, Claude and Cursor.
Build a shared context layer
The Oi Brain page describes persisted knowledge for an organization, its projects and its agents. The connections page describes controlling data access across AI runtimes and using information from connected systems for follow-up work.
Begin with a limited connection and a question whose expected answer you know. Verify which records the assistant can read and whether it cites the information needed for the task. Shared context should not make one teammate's restricted information available to another, so test permissions with the roles you intend to use.
Review workflow actions and guardrails
The homepage gives construction examples involving Procore, MYOB and other systems. A progress-claim workflow combines completion and cost information into material for review. Its example guardrails require referenced scope evidence for variation claims and describe checks against workplace-safety rules.
Those are vendor-described workflows, not a guarantee of legal or safety compliance. Check the actual referenced material and keep qualified review before submitting a claim or acting on regulated guidance. Separate reading records from actions that send messages or change external systems, and confirm where approval is required. A rule's presence in a setup does not establish that every agent will apply it correctly.
Choose a plan by team and message limits
The pricing page lists Hobby for one member, Pro for five and Team for twenty. Hobby includes MCP access and a public library, but lists no agent messages, agent chat or API access. Free access to reusable context is therefore different from a paid allowance for agent activity.
The displayed Pro plan is $19 per month or $190 per year, with 25 agent messages per day and 200 per month. Team lists higher allowances, while advanced governance appears in Enterprise. Confirm how the daily and monthly caps interact and what a trial renews into before connecting a whole organization. Inspect both private and public-library visibility before publishing reusable context that contains internal information.