VITNA – Containment for rogue AI agents
Guard that holds risky MCP tool calls for human approval and signs the decisions
- Pricing
- Open source
- Platforms
- CLI, API, self-hosted
- Runs
- Locally or Remotely
- Version
- 0.5.6
- Listing updated
- 2 days ago
About VITNA – Containment for rogue AI agents
VITNA – Containment for rogue AI agents is an open-source MCP server for security. It can run locally on your machine or as a hosted remote server.
The guard wraps a stdio MCP server you name. Calls to tools marked risky wait for a person, and if nobody answers they do not run. No API key, including the agent's own, can approve a hold. The README's example policy holds write, edit and move calls on the reference filesystem server and allows the rest.
Containment covers only the servers put behind the guard. A built-in shell or an unwrapped server is outside it, and the remote endpoint offers the same checks as tools an agent calls itself, which the project calls cooperative rather than containment. Detection is heuristic pattern matching, not a sandbox.
On a claimed account, decisions are written as Ed25519-signed records that anyone can check offline with a published key and an open-source verifier. With no credentials, the first call self-provisions a restricted trial key, and the claim link is printed in the server log, never sent to the agent. Built-in threat categories include destructive SQL, dangerous shell, secret leaks, prompt injection, exfiltration and suspicious network use, and events export in CEF or LEEF for SIEM tools.
Install
Run in a terminal
claude mcp add --transport http vitna-compliance-preflight https://vitna.costrinity.xyz/api/mcpAdd to Settings → Connectors → Add custom connector
https://vitna.costrinity.xyz/api/mcpAdd to ~/.cursor/mcp.json
{
"mcpServers": {
"vitna-compliance-preflight": {
"url": "https://vitna.costrinity.xyz/api/mcp"
}
}
}Add to .vscode/mcp.json
{
"servers": {
"vitna-compliance-preflight": {
"type": "http",
"url": "https://vitna.costrinity.xyz/api/mcp"
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"vitna-compliance-preflight": {
"serverUrl": "https://vitna.costrinity.xyz/api/mcp"
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"vitna-compliance-preflight": {
"type": "http",
"url": "https://vitna.costrinity.xyz/api/mcp"
}
}
}Configuration
VITNA_OWNER_IDYour VITNA operator UUID. Optional: if unset, the first tool call self-provisions a restricted trial key. The old VIGIL_OWNER_ID name is still accepted.VITNA_API_KEYYour VITNA API key (new keys vitna_..., legacy vigil_... keys remain valid). Optional: self-provisioned if unset. The old VIGIL_API_KEY name is still accepted.VITNA_EMAILEmail to own the self-provisioned trial account. Optional: a throwaway is used if unset. It does not claim the account: claim it with GitHub or a passkey. The old VIGIL_EMAIL name is still accepted.VITNA_BASE_URLVITNA base URL; override only for self-hosted. The old VIGIL_BASE_URL name is still accepted.VITNA_OPEN_CLAIMSet to 1 to have the trial's claim link opened in your default browser when a trial starts. Either way the link is printed in the MCP server log, and the agent is never sent it.
What it does
- Hold risky calls
- Marked tools wait for a person; no answer means no run.
- Signed evidence
- Decision records are Ed25519-signed and verifiable offline.
- Threat detection
- Six built-in categories including prompt injection and secret leaks.
- SIEM export
- Severity-scored events export as CEF or LEEF.
- Policy file
- Lists which actions are allowed and which are held.
- Trial key
- Self-provisions a restricted key on first use.
Good for
- Require approval before an agent writes or moves files
- Block deletions through an MCP server unless a person agrees
- Keep signed records of what an agent was allowed to do
- Feed agent security events into a SIEM
Pricing
Open source
Guard is MIT-licensed; organisation pilot US$7,500 for four weeks
Prices change. Check vitna.costrinity.xyz before you buy.
VITNA – Containment for rogue AI agents FAQ
What does it protect?
Only the stdio MCP servers wrapped by the guard. Anything outside it is not covered.
Can the agent approve its own call?
No. No API key can decide a hold, including the agent's own.
Is the guard open source?
The site says the guard is MIT-licensed.
What does it cost for organisations?
The site lists a four-week pilot at US$7,500.
Is VITNA – Containment for rogue AI agents free?
Yes. VITNA – Containment for rogue AI agents is open source, so you can use and self-host it for free.
How do I add VITNA – Containment for rogue AI agents to Claude Code?
Run this command in your terminal: claude mcp add --transport http vitna-compliance-preflight https://vitna.costrinity.xyz/api/mcp
What are the best alternatives to VITNA – Containment for rogue AI agents?
The closest alternatives to VITNA – Containment for rogue AI agents are REA, Claude Code Ultimate Guide, Agent Roast, Midplane and Leash.
Alternatives to VITNA – Containment for rogue AI agents

REA
MCP serverCLI and MCP server that lets coding agents reverse engineer apps and binaries
Runs locally- Open source
- 464K GitHub stars
- Developer tools

Claude Code Ultimate Guide
MCP serverMCP server for searching the Claude Code Ultimate Guide from any MCP client
Runs locally- Open source
- 6.5K GitHub stars
- AI models and LLM tooling
Agent Roast
MCP serverOffline review that flags agent changes which weaken tests, assertions or CI
Runs locally- Open source
- 4 Hacker News points
- Developer tools

Midplane
MCP serverPostgres gateway for AI agents with table grants, column masking and approval for writes
Runs locally- Open source
- 5 GitHub stars
- Databases

Leash
MCP serverReview what your AI touched, approve risky actions and undo supported changes
- 1 tool
- Remote
- Free
- 1 GitHub star
- Security

SecureCloud
MCP serverSentinelOne and IT security catalog, prices, quotes and checkout links for Israel
- 7 tools
- Remote
- Free
- Security