AI that knows
your company by name.
Sempleo is the team-context layer underneath enterprise AI. Your agents read your systems, your ontology, and your review rules — then finish work and hand you a diff, not a suggestion.
Three layers,
one product.
Your company, modelled.
Sempleo reads your systems of record and maintains a five-layer context model — company, team, client, project, user. Agents read it; humans can edit it.
How context worksWorkers, not chatbots.
Install agents from the marketplace — roughly thirty, across eleven categories. Every one is a readable spec, a rubric, and an audit trail — read it before you run it, fork it when it doesn’t fit.
Browse the marketplaceOpen, by protocol.
Sempleo speaks the Model Context Protocol both ways. Expose your team context to any MCP-aware LLM client; pull in live data from your own MCP servers.
See the platformSits between
your data and your LLM.
Your systems
Data of record stays where it lives. Sempleo reads via OAuth; or pull live data from your own MCP servers. Nothing is copied out without policy.
- Gmail · Drive · Calendar
- HubSpot · Slack · Jira
- Notion · wiki + pages
- Your MCP servers
The context layer
Five-layer context, per-tenant vector store, agent runtime, review queue, and an MCP endpoint — one continuous system, one audit trail.
- Five-layer context model
- Per-tenant vector store
- Agent runtime + audit log
- MCP server · read + suggest
Your model
Prompts go to the provider of your choice, scoped to the minimum context needed. No training on your data — ever.
- Sempleo-managed default
- Bring-your-own Anthropic
- Bring-your-own OpenAI
- Enterprise only
Your systems
Data of record stays where it lives. Sempleo reads via OAuth; or pull live data from your own MCP servers.
- Gmail · Drive · Calendar
- HubSpot · Slack · Jira
- Notion · Your MCP servers
The context layer
Five-layer context, per-tenant vector store, agent runtime, review queue, and an MCP endpoint — one continuous system.
- Five-layer context model
- Per-tenant vector store
- Agent runtime + audit log
Your model
Prompts go to the provider of your choice, scoped to the minimum context needed. No training on your data — ever.
- Sempleo-managed default
- Bring-your-own Anthropic
- Bring-your-own OpenAI
Six of roughly thirty
out of the box.
Email drafter.
Replies to client threads in your voice, grounded in the account history and the last thing they asked for.
In early accessSOW drafter.
Assembles statements of work from the brief, your commercial template, and the client’s prior engagements.
In early accessAccount researcher.
Produces a pre-call brief from CRM, public filings, news, and the notes your team left last time.
In early accessProject status.
Reads Jira, commits, and Slack; writes the weekly status the way your PM writes it — in your house format.
In early accessOnboarding brief.
Packages the kickoff pack for a new hire or new client from the same five layers your team already maintains.
In early accessAnswer desk.
Answers internal questions from your wiki, docs, and threads — with citations, so the answer is checkable.
In early accessFour moves,
kickoff to production.
Connect
OAuth into Gmail, Calendar, Drive, HubSpot, Slack, Jira, and Notion. Point at any MCP server you already run.
Day 1Curate
Seed the five layers from what’s already written down. The founder sits with you while your team reviews.
Week 1Install
Pick agents from the marketplace, fork the ones that don’t quite fit, and wire review queues to your roles. Author your own in plain English when no starter matches.
Week 2–3Operate
Agents land in the review queue with their rationale and citations. Approve, edit, or send back — your team stays in charge.
Week 4“Sempleo exists because I spent years watching teams drown in AI tools that don’t know how their team actually works. We’re building the missing context layer in the open, with our first founding customers.”
