MCP server & API
Your AI assistant, working on the digital twin of your organization
OTS exposes a Model Context Protocol (MCP) server. An authorized assistant can answer an executive's question with figures from the simulation engine, map the current organization, or design a target organization, and every change it proposes is reviewed by a person.
What an assistant can do
- Answer a CEO, CIO or manager with figures: cost, margin, load of each team, bottleneck, absorbable growth, data gaps.
- Run what-if simulations: +25 % demand, AI adoption rate, exception rate, hourly cost, investment.
- Compare scenarios with deltas against the current state: hours freed, payback, ROI.
- Map the existing organization from documents and interviews, as proposals checked against the model's rules.
- Create a target variant and propose a set of organizational changes in it.
- Report what deployed agents and automations really delivered (LIVE).
Connecting
ChatGPT and claude.ai connect from their web versions with OAuth 2.1 (dynamic client registration, PKCE): the user chooses the workspace, the permission level (read, suggest), the data classifications and the duration, and can revoke access at any time. Other MCP clients (Claude Code, IDEs, agent frameworks) use a scoped token. The server address is https://app.otwinstudio.fr/mcp; tokens are created in the application under Settings, AI access.
Governance
The assistant never changes the model directly: proposals wait for a person to review and apply them, objects coming from the AI are flagged as such, and every access and change is logged. The server tells assistants to treat the texts stored in the model as evidence, never as instructions.
The 15 tools
| Tool | What it does |
|---|---|
ots_context | Organization, workspace, scenarios (observed AS IS, TO BE variants) and what the access may do. |
ots_modeling_guide | How to model an organization so that the simulation gives the right figures. |
ots_schema | Exact JSON schemas of objects, relations and proposals. |
ots_read_model | Search and read the model: objects, relations, attributes. |
ots_object_details | One object with all its relations and neighbours: who performs a task, which tools a role uses. |
ots_briefing | Decision briefing computed by the simulation engine, for an executive, a CIO or a manager: cost, margin, load, bottleneck, data gaps. |
ots_simulate | What-if simulation: demand, adoption, exception rate, hourly cost, investment. |
ots_compare_scenarios | Every scenario side by side with deltas against the AS IS: cost, margin, hours freed, payback, ROI. |
ots_impact | What would be affected if a person, role, application or site became unavailable. |
ots_live_status | What deployed agents and automations really did, against the plan. |
ots_analyze | Process durations, waits and explainable automation scores. |
ots_create_scenario | Create a TO BE variant to hold one set of organizational changes. |
ots_validate_changes | Dry-run a plan of changes; errors name the faulty operation. |
ots_propose_changes | Send objects, relations or changes as a proposal that a person reviews and applies. |
ots_list_proposals | Follow the proposals and their review status. |
See your own organization in its twin
Open OTS: explore ten fictional organizations in the live demo, no sign-up needed, or create your own space in a few minutes. The application is available in English.