Built for Serious Work.

Agents That Work Without Supervision

Describe the task in plain English. An agent builds the execution plan and runs it. Pause, resume, or cancel at any point. You get the output; the agent handles the steps in between.

Visual DAG Workflow Canvas

Design multi-step processes as directed acyclic graphs on an interactive canvas. Chain agents, tools, approvals, and conditions into a workflow that is visible, auditable, and runs reliably every time, not just when a human remembers to trigger it.

Scheduled and Recurring Work

Set it once. Run it on schedule. Weekly reports, daily briefings, monthly compliance outputs: generated and routed automatically without anyone manually starting the process.

Built-in RAG and Semantic Search

Index your organisation's documents into a retrieval-augmented generation pipeline that agents query automatically before they act. The right context, surfaced from your own knowledge base, without anyone manually hunting through folders or copy-pasting into a prompt. See how organisations use this for due diligence, compliance, and clinical documentation.

First-Class Workspace and File System

AI that reads, writes, and operates on files inside a cloud workspace you control. Agents create, edit, and share documents as part of a workflow. Everything stays inside your governed environment, not in a chat window you cannot trace or retrieve.

Research and Analysis Workflows

Point agents at your indexed knowledge base, external sources, or both. They gather, synthesise, and return a structured brief, not a raw dump for a human to make sense of.

Apps: Replace Your SaaS Subscriptions

Create complete applications that connect agents, datasets, and tools behind a purpose-built frontend. A compliance tracker, a client portal, an internal research tool: built in minutes, deployed on your infrastructure, talking to each other. One platform instead of a dozen subscriptions.

Persistent, Context-Aware Conversations

Conversations that remember where they left off and what documents they have seen. No re-explaining the context every time. Ongoing work, not one-off interactions.

Guided Model Selection

The system recommends the right model for the task based on capability and cost. Your team spends time on the work, not on evaluating which model to use.

Private Deployment on Your Infrastructure

Run Odokai on your own servers, in your private cloud, in a customer VPC, or fully air-gapped with no external connectivity. Your data does not move. Your infrastructure team stays in control of where it runs. Talk to us about your deployment model.

Exportable Audit Logs

Every model call, every document the AI touches, every step in every workflow: logged with timestamps and exportable in full. Your compliance team can see exactly what happened and when.

Roles and Permissions

Flexible RBAC controls that determine who can access which agents, which models, and which workflows. Clinical staff see clinical tools. Finance sees finance workflows. No one sees what they should not.

Approved Model Catalogue

A centrally managed list of the models your organisation has approved for use. No staff member can route sensitive data to an unapproved model. The catalogue is yours to control and update.

Full Provider Choice, No Lock-In

Use OpenAI, Anthropic, Google Gemini, or open-weight models running locally on your own hardware. Mix cloud and local models in the same workflow: proprietary models where quality demands it, open-weight where data sensitivity or volume calls for it. Swap providers without rebuilding. Onboard a model released today in minutes.

Governance and Spend Controls

Apply rules that control which teams can run which workflows, how much they can spend, and which execution paths are available to them. Governance is not a policy document. It is enforced at the platform level.

Secure Embed Controls

Embed AI capabilities into your own applications with token authentication, origin allowlists, and key rotation. AI that surfaces in your products, on your terms, with full security controls intact.

Complete File Operations

Upload, list, rename, delete, save, monitor, and export files, all inside your governed environment. Nothing leaves the platform without an explicit action that is logged and attributable.

Designed for Security-First Organisations

Opus is engineered around control frameworks aligned with SOC 2 and ISO/IEC 27001 principles, with customer data protection and platform integrity treated as first-order requirements, not afterthoughts.

Controls You Can Take Into Due Diligence

Current controls include authenticated access and role-based permissions, API token lifecycle management, secure defaults (CSRF, strict CORS, rate limiting, and security headers), sensitive-data-aware logging, audit trails for security-relevant actions, runtime guardrails, and monitoring designed for incident detection and investigation.

Our governance, evidence collection, and operational processes continue to mature in support of formal compliance programs.

Human Approval at Every Critical Step

Insert approval checkpoints at any point in a workflow. Agents do the work and surface the decision; a human approves before anything consequential is acted on. The judgment stays with your people.

Real-Time Execution Visibility

Watch workflows execute live, node by node, with progress updates at every step. You are never waiting on a black box. You can see what is happening and intervene if something is wrong.

Cost and Usage Monitoring

Spend and usage data across every user, team, model, and workflow, in one dashboard. No hidden AI costs. No surprises at the end of the month. Budget controls enforced before they are exceeded.

Central Admin Dashboard

Manage users, roles, groups, model registry, feature flags, and operational settings from a single control plane. Your IT and operations teams run the platform; they do not need to delegate that to vendors.

AI Evaluation Framework

Test model and workflow performance before you rely on it operationally. Built-in evaluation tools let you validate outputs, compare models, and iterate before anything goes live.

API Access for Custom Integration

Generate and manage API tokens for secure programmatic access. Connect Odokai to your existing systems without giving up the security controls that govern everything else on the platform.

Full API Access and MCP Integration

The relationship between Odokai and your stack works in both directions. Use the full Odokai API to trigger workflows, manage agents, and pull results into your own systems programmatically. Or connect your external tools into Odokai using MCP (Model Context Protocol), so agents can act on your systems directly. Both directions work. Neither requires replacing what you already have. Read more about delegating work to AI agents.

App and Service Connectors

Browse and activate pre-built connectors for Gmail, Notion, databases, and custom APIs through the connector marketplace. Agents that work alongside your existing tools rather than demanding your team learn something new from scratch.

Third-Party Tool Extensions

Extend what agents can do by connecting external tool servers. Custom capabilities built for your specific processes, without rebuilding the underlying platform infrastructure.

Advanced Orchestration Management

Build and operate complex automated workflows from one control plane. Sequential pipelines, parallel processing, conditional routing, and error handling: the kind of reliability your operations team needs to trust AI with real work.

Team Sharing and Collaboration

Share agents, conversations, and workspaces across users and groups. The expertise built into one agent is available to the whole team, not siloed in a single person's account.

Embedded AI for Your Applications

Surface AI capabilities inside your own products and internal tools using straightforward code snippets. Governed, audited AI, not a public API embedded without oversight.