The agentic layer · Model Context Protocol
Connect your AI agent to ElevatedOS.
Your own AI agent — Claude, and any MCP-compatible client — can read your operating model and propose the next move into your review queue. It can never act, send, or spend. You stay the one who decides.
The boundary
Your agent can read and propose. It can never act.
As agents get more capable, the temptation to let them act unattended grows — and so does the cost of a confident mistake. ElevatedOS draws a hard line, enforced in the server itself, not just promised.
Read and propose only
Agents read defined views and place suggestions in your review queue. No send, no spend, no publish, no external action — ever.
A token you control
Issue a personal token from Settings, hashed at rest and shown once. Revoke it any time. It is the only credential.
Scoped to your business
Every request is filtered to your organization. Cross-tenant access is structurally impossible, not just disallowed.
Fully audited
Every read and every proposal is logged. Sensitive numerics — like strength scores — are never exposed, only labels.
What it can read
Read-only views of your operating model
Resources your agent can pull to reason from a real picture of your business — never a guess.
Founder Operating Model
readThe founder's whole operating state: identity, current week, open delegations, active bottlenecks, cycle, compass, KPIs, team coverage. Strength labels only — no numeric scores.
Founder Dependency Index
readThe Founder Dependency Index (FDI) trend: the latest reading plus the weekly series. Lower is healthier — the business depends less on the founder.
Today's Designed Day
readToday's plan: top focus plus the ordered list of commitments (priorities, delegation follow-ups, anti-priorities) and their state.
Open Delegations
readDelegation items still open or in progress — what the founder has handed off but not yet closed out.
Latest Weekly Brief
readSummary fields of the most recent generated weekly brief for the org (summary, diagnoses, focus plan). Read-only narrative context.
Strength Summary
readThe founder's Strength Finder context — LABELS ONLY (top strengths, lean-aways, dominant domains). Numeric ranks/scores are never exposed.
What it can propose
Suggestions into your review queue
Tools that draft or suggest. Each one lands where you already review and decide — nothing is created, sent, or run automatically.
Mark a daily item done
reversibleMark one of today's daily items as done (or reopen it). Straight-through, low-risk, reversible. Only items in the founder's own org are accepted.
Propose a delegation
proposesPropose a delegation as a PENDING recommendation on the founder's latest brief. It appears in the normal review UI for the founder to approve, edit, or reject — it is never created or sent automatically.
Propose today's plan
proposesCreate or update today's daily plan in 'proposed' status with a suggested top focus. The founder activates it; nothing is executed on their behalf.
Draft this week's reset
proposesSeed this week's weekly reset DRAFT with free-text context and the founder's 'leverage for next week'. Draft only — it never submits the reset or generates a brief. The founder opens it, reviews, and submits.
How to connect
- 01
Issue a token
In ElevatedOS, go to Settings → Agents and issue a personal access token. It is shown once and bound to your organization. Revoke it whenever you like.
- 02
Point your agent at the endpoint
Add ElevatedOS as an MCP server in your agent, using the endpoint below and the token as a bearer credential.
- 03
Your agent reads and proposes; you decide
Ask your agent about your week, your dependency trend, or your open delegations. It can draft a plan or propose a delegation — and every suggestion waits in your review queue for your approval.
MCP endpoint
https://elevatedos.io/api/mcpQuestions
Can I connect Claude or ChatGPT to ElevatedOS?
Yes. ElevatedOS runs a Model Context Protocol (MCP) server. Any MCP-compatible agent — Claude and a growing list of others — can connect with a personal token you issue from Settings, then read your operating model and propose actions into your review queue. The point is leverage without losing control.
Will an AI agent be able to act on my business without me?
No. The server is read-and-propose only. An agent can read defined views of your operating model and place suggestions into your existing human-review queues. It can never send an email, move money, publish, change a system, or take any external action. Every proposal waits for your decision.
Is it safe to connect an AI agent to my business data?
The design assumes it must be. Access is by a token you issue and revoke yourself, hashed at rest and shown once. Every request is scoped to your organization, every read and proposal is logged to an audit trail, and strength scores and other sensitive numerics are never exposed — only labels. The agent can see and suggest; it cannot reach out and change anything.
What is an MCP server?
The Model Context Protocol is an open standard that lets an AI agent connect to an external system in a structured way — listing what it can read (resources) and what it can do (tools). ElevatedOS exposes its operating model as read-only resources and a small set of propose-only tools, so your agent works from a real model of your business rather than guesses.
First, a model worth connecting.
The agent reads your operating model — so it is worth building one. Start with a three-minute diagnostic. No signup required.
Take the diagnosticAlready inside? Issue an agent token.