The ELIZON platform
Build, govern and run AI agents — in one place you control.
Everything an organisation needs to put agents to work: a visual builder, agent teams, company knowledge, live voice, governance and cost control — packaged so you can install the same solution again.
- Email Triggerclaims inbox
- AI AgentClaims triage agent
- Human Reviewpayouts over 5,000
Governance
- Project
- Claims
- Project budget
- €504 used€1,200 / month
- Last run
- Completed · 2.4s · €0.031
Build
From one agent to a team that runs a whole process.
Business teams build in the visual builder. Engineers work on the same objects through the API and the CLI.
Visual builder
Design agents and workflows on a canvas: instructions, tools, knowledge and steps in one view.
Agent teams
Combine agents that route work, coordinate or collaborate on one task.
Workflows
Step-by-step processes with conditions and human approval steps, for work that must run the same way every time.
Company knowledge
Connect documents and data so agents answer from your sources.
Tools and skills
Give agents tools through standard connectors, and reusable skills they can share.
Memory
Agents remember what matters between conversations, within the limits you set.
- Mode
- Route
- Lead
- Intake router
- Member
- Motor claims agent Motor
- Member
- Property claims agent Property
- Member
- Fraud check agent Fraud
Intake router → Property claims agent · 4.2s
Talk
Any agent can take a phone call.
Put an agent on your phone lines, or into your own web app through the voice API, as a live voice agent. It follows the same roles, budgets and audit log as every other agent.
Phone and web
Connect your own phone lines through Twilio, Telnyx, Infobip or any SIP provider, or build voice into your web app with the voice API.
Your choice of voice providers
Swap speech-to-text and text-to-speech providers without rebuilding the agent.
- Agent
- Claims triage agent
- Record calls
- On
- Provider
- Infobip · +49 30 •••• 0142
- Call
- Inbound · Answered · 3m 12s
Voice session · STT seconds 184 · TTS characters 2,310
Govern
Know who can change every agent, and what it did.
Members and roles
Owner, Admin, Editor and Normal, plus custom roles, per organisation and project.
Tools, knowledge and review
Choose the tools and knowledge each agent can use. Workflow steps can wait for a named reviewer.
Traces
Every step of every team and workflow run recorded: inputs, outputs, model, tokens and cost.
Version history
Earlier versions of agents and workflows are kept; compare and publish workflows.
- Assigned to
- Maria Ivanova
- Role
- Claims handler · Agents ✓ view · Knowledge ✓ view · Billing ✗
Audit Log · 09:14 · Maria Ivanova · member.role_changed
- Team · Claims intakeLLM tokens · MonthlyOKCurrent €504Cap €1,200
- Team · Fraud reviewLLM tokens · Monthly80% WarningCurrent €331Cap €400
- Project · Nordhavn pilotVoice minutes · MonthlyExceededCurrent 3,000 minCap 3,000 min
Control cost
Budgets that stop at the cap.
Set budgets per organisation, project or team. ELIZON tracks usage and cost as it happens and stops new work at the cap.
Usage reports break cost down by project and team, and export as CSV or Parquet.
Package
Install a finished solution again — for the next team or client.
Install a signed, versioned package. Each installation connects its own models, tools and knowledge, and can be updated or rolled back.
Partners build packages of agents, teams, workflows, knowledge, tools and test cases with the ELIZON CLI.
How partners use packages →- Version
- 1.3.0 · Signed by Nordhavn Digital
- Objects
- 12 · 3 Agents, 1 Team, 2 Workflows, 1 Knowledge Base…
- Previous
- 1.2.0 Superseded
- Review
- LLM
- Tools
- Knowledge Base
- Validate
- Activate
Run anywhere
Your infrastructure, your models, your tenants.
Deploy where you choose
Your servers, air-gapped, a cloud you choose or your provider’s cloud. Runs on Kubernetes.
Any AI model
Commercial, open-weight or self-hosted. Choose a model per agent and switch without rebuilding.
Many organisations, one install
Separate organisations and projects, each with its own data, members and limits.
Open API
Run agents and read results from your own systems through a public API.
Under the hood
Built for long-running, production work.
Durable workflows
Long-running workflows survive restarts and pick up where they left off.
OpenTelemetry traces
AI worker traces can be sent over OpenTelemetry (OTLP) to your monitoring stack.
Kubernetes-native
Deployed with standard tooling.
Architecture overview
For security and platform teams, sent with the security pack: request it.