Central governance for all things AI

Manage every model, tool and policy in the organization from one place.

Book a demo

Harriet AHadmin@meridianprops.co.uk

Harriet/Usage & spend

Usage & spend

Spend forecast

October · linear estimate

Month to date$8,499

Projected month-end$11,320

Budget$12,000

Daily cost

Group by

29 Sep8 Oct18 Oct28 Oct

Budgets

Organization$8,499 of $12,000

70.8% used · Monthly · alert at 80%
Engineering$3,613 of $4,500

80.3% used · Monthly · soft block at 100%
Sales$2,498 of $3,500

71.4% used · Monthly · alert at 80%

Models your organization can use

Access by group is set on each group

Claude Opus 5Anthropic · EU hosted
GPT-5.5 miniOpenAI · EU hosted
Gemini 3.5 FlashGoogle
Kimi K3Open source · EU hosted
DeepSeek FlashOpen source · EU hosted

Illustrative figures for a made-up organization.

Trusted by leading organizations

What goes through it

Knowledge base

Your documents, policies and guides in one searchable place. Agents, chat and skills answer from them instead of guessing.

Observability

Every call through the gateway is on record, with who made it, which tool ran and what came back. Usage and spend show by skill, model and team.

Integrations

One place for every connector, from Slack and HubSpot to any MCP server. Approved once and given to the right teams, not set up again in every tool.

Take back control of AI in your organization

Spend you can see

Cost by team, model and task, visible as it happens rather than on the invoice.

Universal connectivity

Harriet Desktop, Claude Desktop and cloud workspaces all reach their models through one gateway.

Access by team

Choose which models each team can use, such as the strongest models for engineering and lighter ones for everyone else.

Data protection

EU-hosted versions of leading models and Claude models with zero data retention. A DPA and sub-processor list are ready for your security review.

Session analytics

A full record of sessions and activity per person, so you can explain a cost spike and spot a model used where it should not be.

Questions we get asked

Which models can we use?

Claude, OpenAI, Gemini, Mistral and other leading models, chosen per task. New ones are added as they arrive.

Where is our data processed?

Through the EU for any model, with zero data retention available. The arrangement is set out in our DPA and sub-processor list.

Can we limit which teams use which models?

Yes. Model access is set by group, with SSO and your directory groups as the source of truth.

How do we keep the cost down?

Budgets per team, the bill by model and task as it happens, and eval mode to see which model earns its cost on each job.

Bring one process, we bring the rest

Book a demo