One platform.
Every AI workflow.
mploi is the end-to-end AI operating layer for the enterprise — agent building, knowledge management, secure tool execution, automation, and governance, in a single self-hosted product.
The command center for your AI workforce
Design intelligent agents, equip them with the data and tools they need to act, and deploy them — into chat, into your apps, into scheduled automations — all inside your security boundary.
Compose agents the way you'd compose a team
Every agent is a configurable unit — a name, a personality, an allowed toolset, a knowledge base, and a model. Build them in a visual editor. Save them as templates. Clone them across departments. Spin up an invoice-review specialist one afternoon and a support generalist the next.
When work is too complex for one agent, compose a team: a router delegates sub-tasks to specialists, each scoped to the tools and data appropriate for its role. The platform handles handoffs, context propagation, and result aggregation — you handle the org chart.
Use the right model for every job
mploi treats models as interchangeable. Plug in an API key from any major provider and that model becomes selectable by any agent on the platform. Run open-source models on your own GPUs through an OpenAI-compatible endpoint and assign them per-agent for sensitive data paths.
The same agent can be reconfigured to a different model in a click, with no rewrite. Route low-stakes traffic to lower-cost models, high-stakes traffic to premium ones — or keep regulated workloads entirely on private inference.
AI that knows what your organization knows
Drop in PDFs, Word files, spreadsheets, slide decks, source code, JSON, transcripts, images — even audio and video. mploi indexes everything automatically and makes it semantically searchable to whichever agents you authorize. Agents quote your runbook verbatim, cite your policy document, and reference last quarter's board pack — not the public internet.
Group-scoped knowledge keeps the finance team's docs invisible to marketing's agents. Re-indexing is one click when documents change. Per-user memory carries preferences, decisions, and corrections across sessions.
Agents that take action — within hard boundaries
Out of the box, agents query databases, execute shell and PowerShell commands, transfer files, call REST APIs, browse the web, send notifications, and run a full suite of network-operations primitives. Connect anything else through industry-standard tool servers.
Every tool is permission-gated at the group level. A tool an agent isn't allowed to invoke can't be invoked — full stop, enforced at the platform, independent of what the model decides. Require human confirmation for destructive operations; Plan mode forces a strategy draft before execution.
Set your agents loose on a schedule
Any agent can be scheduled — cron expressions, fixed intervals, one-time future runs. Overnight infrastructure health reports, hourly customer-sentiment digests, end-of-quarter compliance audits. Each scheduled execution runs against a persistent session you can scroll back through later, complete with conversation history and tool-call audit trail.
Pair scheduling with dashboards for real-time, self-refreshing visualizations of whatever your agents collect — uptime, KPIs, queue depth, anything queryable.
Meet your users where they already work
Turn any agent into a chat widget for your internal portal, intranet, customer-facing app, or partner workflow with a single embed snippet. For deeper integrations, expose any agent as a REST endpoint and call it from your existing services, scripts, and back-office systems.
One agent. One source of truth. Everywhere your business needs it.
Agents that hand you finished work
An agent doesn't have to stop at a paragraph of text. It can produce a polished Word report, a multi-sheet Excel workbook, a board-ready PowerPoint deck, or a print-quality PDF — directly from the conversation, with charts, tables, and your organization's visual identity.
For data that's better seen than read, agents render interactive charts and dashboards inline: line, bar, pie, gauges, status boards, network topology — all from natural-language queries against your live systems.
From blank canvas to deployed agent
This is the entire workflow. Pick a model, attach tools, point it at a knowledge base, test it, deploy it. No engineering sprint, no YAML, no waiting — an operator who knows the business can ship a working agent before their coffee goes cold.
Everything the agent can touch is opt-in and group-scoped, so "fast to build" never means "loose to run."
The complete capability set
Everything mploi ships with on day one.
Agent Authoring
- Visual no-code agent builder
- Reusable skills & system prompts
- Per-agent model, tool, and knowledge scope
- Sub-agent chains & supervisor routing
- Plan mode for read-only exploration
- Human-in-the-loop confirmations
Models & Inference
- Connect any commercial provider via API key
- Run open-source models on private GPUs
- Switch models per-agent without rewriting
- Extended thinking / reasoning modes
- Streaming responses with token usage
- Automatic context-window management
Knowledge & Memory
- Documents, code, audio, video, images
- Automatic indexing & semantic search
- Per-source group permissions
- Per-user durable memory
- Auto-extracted facts & preferences
- Continuous memory across sessions
Built-in Tools
- Database query (any SQL backend)
- SSH / PowerShell / shell execution
- SFTP file transfer & REST API calls
- Web browsing & scraping
- Network ops (NETCONF, BGP, SNMP)
- Custom tool servers
Output & Deliverables
- Generate Word, PDF, Excel, PowerPoint
- Inline charts (line, bar, pie, gauge…)
- Self-refreshing dashboards
- Rich Markdown & code rendering
- File attachments in chat
Automation & Deployment
- Cron and interval scheduling
- One-time future runs
- Persistent scheduled sessions
- One-click embed snippet
- REST API for every agent
- Token-based session continuity
Security & Governance
- SSO via OIDC / OAuth
- Group-based access control on everything
- Data-loss prevention with custom patterns
- Tool-level confirmation policies
- Full activity audit log
- Identity provider provisioning
Deployment
- Fully self-hosted, no SaaS dependency
- Single-node or scaled deployment
- Air-gapped operation supported
- Container-based install
- Backup & restore tooling
- Operates entirely in your network
Observability
- Per-agent activity log
- Per-user audit trail
- Tool-call inspection
- Token usage tracking
- Conversation history search
Want to see it running
on your data?
A working demo, tailored to a real workflow from your environment — typically in under a week.
Request a Tailored Demo