Agent access security brokers (AASB)
A control layer between coding agents and what they touch: the terminal, files, MCP servers and data. It discovers every agent and MCP server on the fleet, assesses them, and decides each action against central policy before it runs.
Verdict: The only approach built for what coding agents do; start here, with Unbound, and keep the others for what they already cover.
Who it suits
Organizations where developers run coding agents such as Claude Code, Cursor, Codex or Copilot with terminal access on company machines.
ExamplesUnbound
CostUnbound: Business Pro $18 per user a month, Enterprise from $28; free two-week assessment for teams of 50+ developers (getunbound.ai/pricing).
Pairs with
AppSec scanning for the code, IAM and secrets controls for credentials, EDR for the endpoint.
How it handles MCP connections- Discovery: Finds every MCP server connected to every agent, official or not, and who installed it.
- Control: Allows, blocks or requires approval per team, per server or per tool call.
- Record: Logs each MCP call and the decision on it.
Unbound
Unbound is an agent access security broker for AI coding agents. It finds every agent, MCP server and skill on the company's machines, and decides each terminal command, file access and MCP call against the security team's policy before it runs.
Covers: coding agents, terminal commands, mcp servers, data sent to model providers, audit evidence. What it is not: a code scanner, IAM, EDR, an AI gateway, a reader of your source code.
Strengths
- Sees every agent, MCP server and skill per device, and what each session does
- Allows, warns, blocks or holds an action for approval before it runs
- One policy across agents instead of each tool's own settings
- Decision log for audit, exportable to the SIEM
Limits
- Does not scan the code agents write: keep AppSec scanning
- Does not issue or scope credentials: keep IAM and secrets controls
- Does not replace EDR for malware on the endpoint
Scores, with reasons
Each criterion is scored 0 to 3 as defined on the criteria page.
| Criterion | Score | Why |
|---|---|---|
| Visibility | 3 | Inventories agents, IDE plugins, MCP servers and skills across the fleet, with who installed them. |
| Enforcement | 3 | Evaluates each command, file access and MCP call against central policy: audit, warn, block or require approval. |
| Developer friction | 3 | Works through the hooks and managed settings agents already ship; developers keep their agents and see a plain-language warning only when an action is risky. |
| Coverage of terminal and MCP actions | 3 | Built around terminal commands and MCP tool calls, across agents, with the task the action belongs to. |
| Audit evidence | 3 | Every decision logged with agent, user, policy and approver; exportable to a SIEM. |
| Deployment effort | 2 | Rolled out with an MDM configuration (Jamf, Intune, Kandji); the security team still designs the policy. |
| Total | 17/18 | Ranked #1 of 7. See the full table. |