ATALAIAAI posture
Lists the AI agents and MCP servers your developers and products use, the tools each one exposes and the credentials it can touch.
What it watches
- Agents
Coding agents and assistants on each machine, their version and the folders they can work in.
- MCP servers
Each MCP server: where it came from, the tools it exposes and who configured it.
- Tools and scopes
Shell, file, network and cloud access granted to each agent, and whether it is broader than the task needs.
- Credentials in reach
Tokens and keys an agent can read on the machine or through its tools.
- Prompt-injection paths
Tool descriptions and data sources that can carry instructions into an agent.
- Risks
Agents and servers ranked by what an injected instruction could do with them.
What each agent can reach
A slice of the AI inventory across the fleet. The rows that light up across are the ones that get scoped first.
| Agent or server | Shell | Files outside repo | Network | Cloud credentials | Risk |
|---|---|---|---|---|---|
| Coding agent, default config | ●yes | ●yes | ●yes | ●yes | High |
| Database MCP server | ·no | ·no | ●yes | ●yes | High |
| Browser MCP server | ·no | ●yes | ●yes | ·no | Medium |
| Docs search MCP server | ·no | ·no | ●yes | ·no | Low |
| Coding agent, scoped | ●yes | ·no | ·no | ·no | Low |
How it connects
Every module reports to the tower. Some feed each other.
Reports to
Fed by
The services behind it
The module watches. These services set the station up, with your team, so there is something worth watching.
In development. Early-access teams get it first and shape what it watches.
Join early access →