Meridian features
Nineteen building blocks, one platform.
Everything your people and agents work with, in four groups. Each one opens to how it works.
Build
What you compose.
- 01
Agents
Each agent is a member of the workspace with its own access, a person who answers for it, and memory you can read.
- 02
Skills
Your methods, written once as packages an agent opens when the work calls for them.
- 03
Tools
Every action an agent can take is a tool, and every tool call passes the same policy before it runs.
- 04
Routines
Work that starts itself: on a schedule, on an event or when a row changes.
- 05
Apps
Interfaces on your data and agents, built in a conversation and deployed as real code.
- 06
Databases
Real PostgreSQL databases your agents and your people work in side by side.
Know
What agents can see.
- 07
Company brain
One index over everything you connect, searched under the permissions of whoever is asking.
- 08
AI search
Files from Drive, OneDrive and SharePoint, found by what they say rather than what they are called.
- 09
Connectors
Your systems become tools: connected once, used under a policy you set per action.
Work
Where people meet agents.
- 10
Spaces
A Space gives a team, a project or a client one audience for everything created inside it.
- 11
Channels
The same agents answer in the web app, in Slack, by email, on WhatsApp and from any MCP client.
- 12
Inbox and approvals
Customer conversations, tickets and approvals arrive in one queue, ordered by what needs a person.
- 13
Office
Meetings come in as transcripts and recaps. Documents go out in your brand.
- 14
Devices
Native clients for web, iPhone and Mac over one backend, so work continues where you left it.
Run
How it stays under control.
- 15
Sandboxes
Every chat, app and member gets isolated compute, rebuilt from the database whenever it is needed.
- 16
Models
Every model call goes through one gateway, with failover between providers and limits you set.
- 17
Security and access
Access is decided once, from groups and grants, and enforced by the database on every query.
- 18
Control
Everything an agent does is a Run: recorded, costed and scored.
- 19
Simulation and evals
Test an agent against simulated users and tools before it touches anything real.

Next step
See it on your own data.
Two hours with your leadership. Ten days to a business case. Four weeks to a working agent on your data. No commitment.