Prismor

Prismor is an AI agent security control plane. It sits between coding agents and the actions they take, intercepting every tool call, enforcing security policies and keeping a full audit trail. It is built for enterprises that need to control the identity, access and runtime actions of agents such as Claude Code, Cursor, Codex, Windsurf and Copilot. The open source runtime handles command interception and policy enforcement, secret masking and cloaking so an agent works with placeholders instead of real keys, prompt injection detection, and a tamper evident audit trail stored locally. The hosted platform adds a live dashboard of agent sessions and tool calls, an MCP gateway that puts one policy layer in front of every MCP server, an MCP Hub for seeing and denying servers and tools, contextual rules that understand sessions and repositories, a skill checker that scans agent skills for hidden risks, and 30 days of session history. The Free tier is the open source CLI and runtime at $0, with no account needed and everything running on the user's machine. Starter is $15 per month for one user and up to 5 agents, with 100,000 tool interception events and 1,000 skill scans per month, and the first month is free. Enterprise pricing is custom and adds unlimited usage, SSO and SCIM, open telemetry routing to a SIEM, human approval per policy and on premises deployment.