Operate module · Generally available

Operate your AI agents
with Workbench

Workbench is the operations module of the Matimo platform: production deployment of governed AI agents at enterprise scale - multi-tenant, auditable, and built for the governance requirements your legal team will actually approve.

Multi-tenant isolation
Full execution audit trail
Governance built in
16 reasoning strategies - selectable per agent
Matimo Workbench / Studio: Customer Onboarding Agent v2.1 EXAMPLE WORKFLOW
Node Library
⚡Trigger
⇄Router
◈LLM Engine
⚙Tool Call
↗Output
Governance
🛡 Matimo Policy
📋 Audit Log
⚡
Webhook Trigger
POST /v1/agent/onboard · Tenant: acme-corp
DONE
◈
Intent Classifier - GPT-4o
Engine: reasoning-v3 · Tokens: 312 / 4096
RUNNING
⚙
CRM Enrichment Tool
HubSpot API · Governed by: policy/crm-read-only
QUEUED
↗
Structured Output + Notify
Slack + Email · Schema: OnboardingResult v2
QUEUED
Execution Trace
StatusRunning
Latency1.24s
Tokens used312
Cost$0.0014
Tenantacme-corp
AuditLogged ✓
Governance
🛡 Matimo Active policy/crm-read-only
policy/pii-redact-v2
2 policies enforced
Matimo Workbench in 30 seconds

Enterprise AI deployment
is broken at the seams

Most AI orchestration tools are built for demos or developer experiments. When enterprises try to productionise them, three walls appear immediately.

Governance is bolted on, not built in

Policy enforcement, tool access controls, and audit trails are afterthoughts - left to teams to wire together manually after the system is already running.

Multi-tenancy breaks at scale

Self-hosted solutions require every tenant to be a separate deployment. Data leaks between contexts. Operational overhead compounds fast.

Infrastructure costs are unpredictable

Always-on agent infrastructure costs money even when nothing is running. Per-seat licensing punishes growth. There's no model aligned with actual usage.

Everything enterprise deployment
actually requires

Matimo Workbench is not a framework. It's a production system, designed for the constraints that enterprises actually operate under.

16 Reasoning Engines - Choose the Right Strategy

Matimo Workbench supports 16 distinct reasoning strategies: from ReAct and Chain-of-Thought to Tree-of-Thoughts, Plan-and-Execute, Reflexion, and Multi-Agent coordination. Select the right engine for each agent based on your task profile: structured reasoning, iterative refinement, goal decomposition, or multi-step planning.

OpenAI Anthropic Google Ollama Multi-LLM routing

Matimo Studio

Operations teams compose multi-step agent workflows on a drag-and-drop canvas. Engineers extend and integrate via REST API. Both interact with the same execution engine - no duplication of logic.

Drag-and-drop canvas REST API
Explore Matimo Studio →

Knowledge Bases and Memory

Upload PDFs, websites and documents into knowledge bases with RAG retrieval, so agents recall by meaning, not keyword. Session memory and long-term semantic memory let agents remember across conversations.

RAG Semantic memory

Full Execution Audit Trail

Every token, every tool call, every decision - logged with tenant context and policy state. Exportable for SIEM integration. Built to support compliance requirements as they mature.

Per-tenant logging Exportable logs

Native Multi-Tenant Isolation

Tenant context is enforced at the execution layer - not as application logic. Data, tool access, and execution budgets are isolated by default.

Tenant scoping Context isolation

Human-in-the-Loop Approval

Agents pause on high-risk tool calls and wait for a human to approve or reject before proceeding. The approval queue is built into the admin dashboard - no deployment required to gate any action.

HITL gate Matimo policy engine

Tool Providers + MCP Extensibility

10 built-in Matimo tool providers - Slack, Gmail, GitHub, Notion, HubSpot, PostgreSQL, Mailchimp, Twilio, Microsoft, and Bruno - with 129 individual tool actions, each governed by Matimo's policy engine, plus a governed 449-tool Composio catalog. Connect any MCP server for unlimited extensibility.

129 tool actions MCP compatible

Stateful Agent Memory

Agents maintain context across multi-step workflows and long-running processes. Memory is tenant-scoped, TTL-managed, and governed - not a global mutable store.

Long-running Tenant-scoped

Operations Control Plane

A real-time admin dashboard for your AI operations team. Monitor live executions, trace failures, review usage costs by workflow and tenant, and manage HITL approval queues - all without touching the backend. The visibility your ops team needs once agents are running in production.

Live monitoring Failure tracing Cost attribution HITL approvals

Operate, governed
by the rest of the platform.

Workbench is architected around Matimo OSS, the open-source tool-execution and governance SDK the platform is built on. The platform is designed so that policy enforcement, tool access controls, and audit trails are structural - not bolted on after deployment.

For enterprises that need centralised policy management, SAML SSO, compliance evidence reports, and a full agent identity layer, Matimo Governance is available as an add-on, extending Matimo Workbench's governance capabilities with Shadow Mode, Emergency Stop, Agent Soul documents, and compliance evidence reports across six frameworks. Gateway and the open-source AGDK extend the same governance to agents built outside Workbench.

Explore Matimo OSS →
Matimo Workbench
Operate: Workbench
Workflow engine · Multi-tenant runtime · Cost tracking
▼ governed by
Govern and Open Source: Gateway, Governance, OSS
Policy registry · Tool RBAC · HITL approval · Audit trail
▼ runs on
⚙️
Enterprise Runtime
Cloud-hosted on AWS · Multi-tenant ECS Fargate · RDS PostgreSQL

Built for the teams that
run the enterprise

Matimo Workbench is designed for three distinct roles; each gets what they need without stepping on the other.

AI / Automation Teams

Design agent workflows that actually ship

Use Matimo Studio to compose multi-step agent workflows, connect to enterprise systems, and deploy to production - without managing infrastructure or fighting governance requirements.

Engineering Teams

Extend via API without reinventing the stack

Full REST API access gives engineers control over agent configuration, workflow execution, and tool management - while inheriting multi-tenancy, audit trails, and governance by default. Skip the boilerplate. Build what actually matters.

IT / Security / Compliance

Control what runs, what it touches, and why

Matimo's policy engine classifies tool risk automatically - low-risk actions execute freely, high-risk actions route to a human approval queue. Every execution is logged with tenant context. Audit logs exportable for your SIEM.

Built for production
from day one

Key architectural capabilities designed around what enterprise deployments actually require.

16
Reasoning strategies - ReAct, CoT, ToT, Plan-Execute, Reflexion, Multi-Agent, and more. Select the right one per agent.
1 platform
Native multi-tenant isolation: tenants are separated by default, with no separate deployment per customer or business unit
5 LLM providers
OpenAI, Anthropic, Google, Ollama and AWS Bedrock, with no lock-in. Every execution is logged with tenant context
129 tools
Tool actions across 10 Matimo providers - Slack, Gmail, GitHub, Notion, HubSpot, PostgreSQL, Mailchimp, Twilio, Microsoft, Bruno. Plus a governed 449-tool Composio catalog and any MCP server.
Operate module · Generally available

Ready to run your
first governed agent?

Workbench is generally available. Sign up directly and start building, or schedule a walkthrough with the team if you'd like to see it in action first.

Building on Matimo OSS already? Your policies carry over.