Self-hosted · Version 4.94
Give ErezAgent a business goal. It turns it into strategy, missions and governed execution across your tools, your team and your customers. Nothing consequential happens without your approval.
The Command Center
The Command Center is a live map of your organisation. Every node is a real agent, grouped and coloured by department, with the ErezAgent neural core at the centre. It is where you see what is happening right now.
Hover a node for its name, role, department, status and trust level. A node pulses faster and larger while its agent has work.
Lines show the real relationships: an agent that owns an active mission, and a delegation between two agents as it happens.
Click a node for its profile, current mission and recent delegations, and jump back to the conversation that started the work.
Switch to a flat view with a department legend, or pick a venture to see only its activity.
Pending approvals have Approve, Reject and Discuss buttons. Live operations, a CEO snapshot and your cost today sit on the same page.
An agent with a configured name and avatar shows it here. Identity only changes how an agent is shown, never what it may do.
Watch
Narrated films with captions. Start with one minute, or go straight to the demo.
4 minSee it workProduct demoA small design studio takes one sentence, “my first ten customers in ninety days”, through a goal brief, a Board debate and an approval.
1 minStart hereErezAgent in one minuteThe idea in sixty seconds: a goal in, governed work out, and you in charge of every consequential step.
3 minFounders and business ownersWhat it does for a businessThree kinds of business, from a goal with no plan to an operation already running, and what ErezAgent does for each.All films, including the technical view and the partner briefing
What it is
01
Agents act only through named skills and connectors, and every action passes the same gate: role, risk, autonomy tier, spending limit and approval. The model can reason about a change; that does not give it permission to make one.
02
An Executive Board of AI roles deliberates on your goal, challenges itself, and can abstain or ask for missing evidence instead of manufacturing certainty. Approved direction becomes a measurable objective with stated assumptions.
03
The application and its primary data stores run on infrastructure you control, installed with a guided installer on Windows, macOS or Linux. Connected cloud services receive only what each integration needs.
Start where you are
ErezAgent does not need a finished plan. It meets you at the stage you are at and moves you to a measurable goal, with every step proposed before anything is created.
Idea to first customers
Type or say what you want to achieve. ErezAgent shapes it into a goal with a metric, a target and a timeframe, and shows you the drafts it implies before it creates a single record.
Find the opportunity
Describe the outcome you need, such as a monthly revenue target, and ask for opportunities. ErezAgent returns a ranked shortlist of real options and keeps near-duplicates out of it.
Existing business
Upload files or describe the business. ErezAgent proposes an evidence-backed model, you review and approve it, and only then does it configure the platform around how you already work.
Agencies, MSPs and cloud providers
Deliver ErezAgent-run work to clients from separate workspaces, or deploy a dedicated installation per client and see its status from one fleet view, with the customer’s consent.
Executive Office
Executive Office is where you state goals, ask questions and launch work, by typing or by voice. Everything it starts is tracked against an outcome and waits at the same approval gate.
Chat, the command bar and voice all run through the same assistant, so a conversation resumes wherever you pick it back up. A goal stated in plain words becomes a brief, not a surprise.
What needs you, what moved and what is at risk, assembled from real signals across the company instead of a dashboard you have to read.
Sending, publishing, deploying, deleting and spending carry a risk level and an autonomy tier. Approve or reject in the app, or from Telegram if you connect it.
Signals from across the company are ranked by impact and urgency, so a pipeline below threshold surfaces before it becomes a quarter.
Forty-one specialised agents sit under executive roles. People and agents can own the same tasks, and a step that needs a missing capability is flagged before the mission runs.
Each agent execution is logged with its risk label, token count and time, so you can answer what happened, who approved it and what it cost.
Voice
Executive Conversation Mode is hands-free. You speak, ErezAgent transcribes, answers aloud and starts listening again. The same approvals still apply to anything you ask it to do.
Listening
Continuous turn-taking with an adjustable silence threshold. The original press-to-talk mic still works for a single command.
Barge-in lets you talk over a reply. It is off until you turn it on.
Speech-to-text detects the language you speak instead of assuming English.
When a spoken request ends in “here is the page”, ErezAgent opens it for you.
An active voice session can speak an urgent alert unprompted, only if you enable it.
OpenAI or ElevenLabs, or your own cloned voice on a local voice server you run yourself. Agents can carry a distinct voice, and a voice never grants authority.
The loop
Every objective names the value it should create, how it will be measured and which assumptions it depends on. Results return as evidence for the next decision.
How it works
Five steps, one continuous record. Scroll through a real run, from a stated goal to an approved action and the evidence it left behind.
01 Goal
Type or say what you want to achieve. ErezAgent shapes it into a goal brief with a metric, a target, a timeframe and a scope, then offers the drafts it implies: a company priority, a venture or an objective.
Nothing is created until you confirm it.
02 Board
The Executive Board brings company and venture context, history and current evidence to the goal. Each role answers on its own, a challenge round surfaces disagreement, and the synthesis lists the evidence still missing.
Where the evidence is thin, the Board says so.
03 Work
Missions are planned against the capabilities that are actually configured. Forty-one specialised agents carry out the work through 260+ named skills, and people and agents can own the same tasks.
A step that needs a missing capability is flagged before the mission runs.
04 Approval
Sending, publishing, deploying, deleting and spending carry a risk level and an autonomy tier. Low-risk work can run on its own; higher-risk actions create an approval; the most critical never run autonomously.
You approve or reject in the app, or from Telegram if it is connected.
05 Evidence
State changes are written to an audit log that the application cannot edit. Cost is measured per agent, model and mission, and results flow back to the objective they were meant to move.
You can see what happened, who approved it and what it cost.
Platform
The same governed execution path serves chat, missions, scheduled jobs, connectors and MCP. Open any area for the detail.
Multi-role deliberation with a challenge round, abstention and requests for missing evidence.
Details41 specialised agents with their own tools, budgets, models and policies per workspace.
DetailsMeasurable objectives with metrics, targets and falsifiable assumptions, driving planned missions.
DetailsAn evidence-backed model of an existing business, applied only after the owner approves it.
DetailsCustomer workspace, document collection, support, customer success and appointments.
Details39 managed connectors, a 42-provider catalog of API actions, and governed inbound and outbound MCP.
DetailsAutonomy tiers T0 to T3, approval gates, spending ceilings and an audit log.
DetailsA multi-container stack with a guided installer, health checks, upgrades and rollback.
DetailsTrust
The language model reasons about what should happen. The execution layer decides whether it may happen now, must wait for approval, or is blocked.
Governance and security in detailPlain about the edges
Plans for how you work
ErezAgent is licensed for self-hosting. Plans differ mainly by capacity: ventures, active projects, users and client workspaces. Optional add-ons are priced separately, and you bring your own AI-provider usage.
One person, one venture to start. Move from a stated goal to an approved plan and the first missions without hiring a team to run the software.
$99 / month
or $990 per year
$249 / month
or $2,490 per year
Bring your documents and processes. The Business Twin proposes how you operate, and ErezAgent configures records, workflows, rules and CRM around it, with your approval.
$999 / month
or $9,990 per year
$1,999 / month
or $19,990 per year
Give each client its own workspace and run their growth, support or operations through governed agents, with an audit trail you can show them.
$599 / month
or $5,990 per year
$999 / month
or $9,990 per year
Deploy a dedicated installation per client, or run client workspaces on one. License downstream installs with your own signing key and see their status from one place.
Custom
Priced to your model
Prices in USD for a self-hosted licence. Annual billing costs 10 months. Paid evaluation and guided deployment: talk to us. Compare every plan
Built by
ErezAgent was created from the belief that artificial intelligence should operate as an intelligent company rather than as isolated assistants.
Next step
See it run on a goal of yours, apply for early access, or talk to us about delivering it to your own clients.
Paid evaluation and guided deployment: talk to us. See the plans.
