Straight answers on positioning, ownership, and adoption.

A harness is the system that wraps a model and makes it useful at work. Context loading. Tool orchestration. Error handling. Output routing. Integration with the systems where decisions get made. The model is the engine. The harness is the chassis, steering, dashboard, and brakes. Cursor proved it for developers: the same Claude model scores 80.9% on SWE-bench inside Cursor's harness and 45.9% on a bare scaffold. A 35-point swing from harness craft alone. Odokai is the equivalent harness for the rest of the office.

Because developers got tools built around their workflow. IDE-native, integrated with version control, aware of the codebase, governed through existing engineering practice. The rest of the workforce got a blank text box and a vague instruction to make it work. Compliance officers, project managers, analysts, legal teams, and operations leads have no equivalent. So they copy-paste sensitive data into consumer AI tools, or quietly avoid AI altogether. Both outcomes are bad for the organisation. Odokai exists to close that gap.

Same product category, different model. The hosted incumbents give your team a consumer-grade assistant with an enterprise login. Odokai gives you a harness: multi-model chat, custom agents, knowledge with RAG, workflow orchestration, apps for teams. Plus the surface those tools don't have. Full extensibility. Multi-vendor model choice. Deep governance. Customer-owned deployment. Hosted products make sense when convenience is the priority. Odokai makes sense when ownership, control, and the ability to extend matter more.

Three things. First, the harness deploys into infrastructure you control. Your AWS, GCP, or Azure tenant, your on-prem environment, or fully air-gapped. Your data does not have to leave it. Second, every agent, workflow, knowledge base, and app your team builds belongs to you. Prompt and tool configurations included. Third, the commercial relationship is built so that if you ever wanted to take the platform code and run it yourselves indefinitely, you could. We don't hold your stack hostage to keep you a customer.

Yes. Governance is in the architecture, not bolted on. RBAC across users, groups, and roles. Action-level audit logging that is traceable and attributable. Policy guardrails that block sensitive-data leakage and enforce regulatory requirements. Human approval gates for high-risk actions. An evaluation framework for benchmarking agents before rollout. The platform is designed to be SMCR-ready for environments where Senior Managers are personally accountable, and is aligned with the EU AI Act's risk-classification, transparency, and human-oversight requirements from day one. Combine that with private-cloud or air-gapped deployment and Odokai meets the operating bar of financial services, healthcare, government, and other tightly regulated sectors.

Shadow AI is AI usage happening outside IT oversight. It usually starts with good intent, teams trying to move faster with personal ChatGPT tabs, browser extensions, or ungoverned tools. The risk is operational, legal, and reputational: no audit trail, no policy enforcement, and no reliable control over sensitive data.

Odokai gives teams the same speed in a controlled environment. Every model call can be policy-scoped, approval gates can be added to high-risk steps, and activity is logged so governance teams can see what happened and why.

Yes. A single gateway sits in front of supported providers. Route by team, by workflow, or by data sensitivity. Run open models like Llama or Mistral via vLLM or your inference of choice. Plug in fine-tuned models you have trained yourselves. The harness handles model-specific tool formatting and context strategy so each model operates at its best. When a better model ships, you swap the model. The workflows, prompts, approvals, and apps on top do not change.

There aren't any closed walls in the surfaces that matter. The harness is MCP-native, so any Model Context Protocol server you can write or buy plugs straight in. Custom tools are written in JavaScript or TypeScript with first-class testing. The connector layer ships declarative integration packs for the systems you already run: CRM, document management, case management, ERP, the databases behind them. Studio lets you build app UIs on top of agents with live preview, embeddable in your product. Everything is also available through a versioned API and webhook surface. The "describe it, build it" experience Lovable proved for app code, applied to office workflows.

Depends on the deployment mode and how much you want to build. Managed needs no infrastructure work on your side. Private cloud and air-gapped need an engineer to operate the platform the way they would any internal service. Building agents, workflows, and apps inside the harness is designed for product-minded operators, not only engineers. Engineers can extend it deeply when you want them to.

Managed deployment can have the first agent, knowledge base, and team operational in days. Private cloud deployments typically complete initial setup within two to four weeks. The first production workflow with real integrations and approvals usually lands inside four weeks. Speed is gated by how quickly your team can confirm the workflow logic and sign off on what systems agents can touch. Not by platform complexity.

Two patterns are common. For many organisations, Odokai replaces the per-seat ChatGPT or Copilot sprawl with a single governed harness where the same models are available behind your policy. The data, the prompts, and the institutional knowledge stay yours. For organisations where leadership wants to keep individual ChatGPT or Copilot for personal productivity, Odokai becomes the platform for everything that touches real systems, real data, or real business workflows. Both work.

A Team plan at £99/user/month consolidates the spend a company would otherwise make on separate tools — CRM, project management, content planning, compliance tracking, and reporting dashboards. Build equivalent apps inside Odokai Studio and they live on your infrastructure, governed by your policy, without paying per-seat fees to a separate vendor for each one.

Odokai offers four plans: Pilot, Team, Enterprise: Cloud, and Enterprise. Pilot is the managed starting point with 1,000 monthly credits. Team is built for governed rollout with 250+ models, admin controls, API access, and 5,000 monthly credits per user.

Enterprise: Cloud deploys Odokai into your own private cloud account. Enterprise covers on-prem and air-gapped deployment with dedicated support and BYOT. The practical path for most teams is Pilot to Team, then Enterprise deployment when infrastructure ownership or isolation becomes a requirement.

Credits are the usage unit across models, agents, and workflows. Higher-capability models consume credits faster. More efficient models consume credits more slowly. This gives you clear control over quality and spend per task.

In practice, teams use a mixed routing strategy: efficient models for bulk and routine work, higher-reasoning models for complex decisions. One pricing system, one governance layer, no vendor lock-in.

It starts with one process. Say your compliance team currently relies on an external consultant to review documents and produce quarterly reports. You build an agent in Odokai that does the first-pass analysis. You build an app in Studio that the compliance team uses to review exceptions and sign off. The agent, the workflow, and the app all live inside your Odokai environment, governed by your policy, with full audit trail. Your team owns the capability. Repeat this pattern across marketing, sales, operations, legal. One by one, external dependencies become internal capabilities built on the harness.

Yes. Studio is the app builder inside Odokai. You describe what you need: custom fields, data views, workflows, access controls. Studio generates the UI and connects it to the agents and knowledge bases behind it. It runs inside your Odokai environment, on your infrastructure. Your data, your fields, your users. No per-seat SaaS pricing for yet another tool. The app is a feature of your existing Odokai subscription. If your needs outgrow Studio, you can embed Odokai surfaces into your own frontend via the API.

Odokai combines six surfaces in one platform: chat, agents, workflows, knowledge (RAG), Studio apps, and connectors. Chat handles day-to-day assistant work. Agents and workflows handle repeatable operations. Knowledge keeps answers grounded in your documents. Studio and connectors let you build and integrate the operational layer around all of it.

This matters because it replaces a fragmented stack with one governed system. Teams spend less time stitching tools together and more time shipping outcomes.

There are three deployment modes: managed hosting by Odokai, deployment into your own cloud account, or on-prem and air-gapped installation. The product surfaces stay consistent across modes, so moving later does not require rebuilding workflows and agents.

Most teams start managed for speed, then move to private cloud or isolated environments when policy, procurement, or sovereignty requirements demand tighter control.

Odokai supports delivery at multiple levels, from a focused discovery conversation through to full production implementation. The usual progression is: identify the highest-value workflow, run an opportunity sprint to validate feasibility and ROI, then ship a production build on your real data and systems.

For teams that want ongoing capability growth, we also provide fractional AI leadership and embedded delivery support. Each engagement is designed to leave your team stronger, with documentation, operating guidance, and a clear handover path.

Start with one workflow where delays and manual effort are obvious. Good first candidates are compliance checks, recurring reporting, support triage, and account research. We map the current process, define approvals, set policy guardrails, connect your systems, and launch a first production workflow.

Once the first workflow is live, expansion gets easier. You already have governance, model routing, and operating patterns in place, so the second and third workflows land faster.

All plans include product documentation and baseline support. Team plans include priority support for operational usage and admin management. Enterprise plans include dedicated support, guided onboarding, and implementation partnership for private-cloud, on-prem, and air-gapped deployments.

The support model is built around production reliability, not just ticket response. That means faster resolution paths, clearer ownership, and practical guidance as your usage expands.

The best first use case is repeatable, high-friction, and owned by a specific team. Typical examples include first-pass compliance review, monthly management reporting, sales account preparation, and knowledge-grounded support drafting.

The goal is simple: prove operational value quickly while approvals, traceability, and governance are already built into the process.

Yes. The Team plan gives you the full Odokai platform: chat, agents, workflows, knowledge with RAG, Studio apps, MCP-native connectors, 250+ models, 100+ tools, RBAC, audit, admin dashboard, and API access. 5,000 credits per user per month. Minimum 5 users, 12-month subscription. What it does not include is the managed deployment to your own private cloud or air-gapped environment — those are Enterprise plans.