Our pricing
Testing your agent is available today, at a fixed price or on quote. The Shield, the software that blocks unauthorized actions, is in development.
Red: the test (available)
We test your agent the way an attacker would and show you what gets through today.
Your agent reads tickets, documents and messages. We hide instructions in them, the way an attacker would, on your test environment and with written authorization. Every action it takes that it shouldn't is documented.
A classic pentest usually costs between €5,000 and €25,000 and takes several weeks, without testing your AI agent. The Agent Red Team Express starts within 48h and delivers in 5 days. For a wider scope or an ongoing collaboration, we put together a quote.
Agent Red Team Express
€1,900
Ideal for : Finding out quickly what a stranger can make an agent do.
- What your agent can do: actual access vs. needed access
- Reproducible demonstrations: one booby-trapped text, and the agent complies
- One documented data-leak attempt
- Report you can reuse in customer questionnaires
- 30-minute debrief with your team
- Report delivered even if nothing gets through: what held up is documented
Custom Red Team
On quote
Ideal for : Several agents, an agent with many tools or MCP servers, or an agency delivering agents to its clients.
- Scope defined together: agents, tools, MCP servers, environments
- For agencies: a red team before each client delivery
- White-label option, report in your agency's format
- Re-test after fixes
- One-off engagement or ongoing support
Blue: the Shield (in development)
Software that sits between your agent and its tools. Allowed actions go through, everything else is blocked.
The Shield sits between your agent and its tools: a list of allowed actions, per-action limits, flow rules and a decision log. Rules are versioned in your repository, and the software runs on your side.
Early customers will get access first, in exchange for feedback. Every testing engagement helps refine its rules. Leave your details to hear when it launches.
The Shield
In development
Ideal for : Keeping your agents' actions under control over time, with readable rules.
- List of allowed actions: everything else is refused
- Per-action limits: scope, amounts, paths, recipients
- Flow rules: sensitive data that was read can't leave through an outbound action
- Searchable decision log, versioned rules
- Runs on your side, with direct support for early customers
Purple: the test, then the Shield
Testing and protection in a single engagement.
The test lets us write your agent's rules now. The Shield will enforce them on your side as soon as it launches, with priority access and locked-in pricing.
The Shield isn't available yet, so the Purple program opens at launch. Signing up now reserves your spot and your price.
Purple Program
At Shield launch
Ideal for : A critical agent, where you want both the evidence and the control.
- Red team of your agent, then Shield installation at launch
- Rules written for your stack
- Priority access to the Shield
- Locked-in pricing
- Debrief and follow-up with your team