AI agents are out of control.
Gaincontrol finds the AI agents in your code and governs what they're allowed to do: Quin inventories them, and Zino authorizes each call before it runs.
Simulated fleet
Discover. Authorize. Control.
Three focused products. Each works on its own; together they cover discovery, identity, and execution for your AI agents.
Quin
Finds the AI agents in your repositories, with the prompts, tools and models each one uses.Find agents before they become blind spots.$ quin-scanner scan ./payments3 agents found · 1 risky promptZino
Checks every agent tool, network and MCP call against policy before it runs.Every agent earns its access, nothing more.refund-agent → payments.refunddenied · no grant for this call · loggedDrona
Early accessBeing built to check each agent action against a workflow contract before it runs.AI is probabilistic. Execution isn't.intent → contract → stepsin private developmentThe scale of the problem
Each figure feeds the next: agents run that nobody registered, they hold more access than they need, and the bill lands after the breach.
82%
of organizations found an AI agent or workflow that security/IT didn't know existed
Quin finds them in your repositoriesCloud Security Alliance & Token Security, 202697%
of non-human identities carry excessive privileges
Zino scopes each agent's identityEntro Security, 2025 State of NHI Report$4.63M
average cost of a breach involving shadow AI, vs. $4.44M baseline
Zino can revoke a rogue agent in secondsIBM Cost of a Data Breach, 2025
Agentic AI incidents
- Jul 2026Misuse & escapeClaude models in cyber evaluations reached three real organizationsAnthropic(opens the source in a new tab)
- May 2026Supply chainMalicious ClawHub skills used OpenClaw agents to install infostealersUnit 42(opens the source in a new tab)
- Apr 2026Data destroyedCursor agent deleted a startup's production database and its backupsDecrypt(opens the source in a new tab)
- Feb 2026Data destroyedClaude Code ran terraform destroy, deleting a production databaseAuthor's postmortem(opens the source in a new tab)
- Feb 2026Prompt injectionPrompt injection in Cline's triage bot led to an unauthorized npm releaseCline postmortem(opens the source in a new tab)
- Dec 2025Data destroyedGoogle Antigravity agent wiped a user's entire D: driveThe Register(opens the source in a new tab)
- Nov 2025Misuse & escapeState-backed group used Claude Code to run most of an espionage campaignAnthropic(opens the source in a new tab)
- Sep 2025Supply chainFake postmark-mcp server BCC'd every agent-sent email to an attackerCSO Online(opens the source in a new tab)
- Aug 2025Supply chainMalicious Nx packages drove local AI coding CLIs to hunt for secretsStepSecurity(opens the source in a new tab)
- Jul 2025Data destroyedReplit agent deleted a database during an explicit code freezeThe Register(opens the source in a new tab)
- Jul 2025Data destroyedGemini CLI lost a user's files during a failed file-move sequenceAI Incident Database(opens the source in a new tab)
- Jul 2025Supply chainAmazon Q extension shipped with an injected prompt to wipe systemsThe Register(opens the source in a new tab)
- Jun 2025Prompt injectionMicrosoft 365 Copilot flaw let a single email trigger data exfiltrationCybersecurity Dive(opens the source in a new tab)
A system prompt is advice. A permission is a control.
In April 2026 a coding agent working in a startup's staging environment found an account-wide API token in an unrelated file and used it to delete the production database, backups included, in nine seconds. Here is what happened, beat by beat, and the control that applies at each step.
Sources: The Register, Decrypt, Railway
Apr 24
An agent finds a root key
A Cursor agent was doing a routine task in PocketOS's staging environment. In a file unrelated to that task it found a Railway API token created only to manage custom domains, but provisioned with account-wide access. As founder Jer Crane put it: "Didn't give it access, it found it."
Inventory: know what the agent can reach
Quin- What happened
- Nobody had a record that a long-lived, account-wide token sat where the agent could read it.
- With the control in place
- The agent, and the credentials, endpoints and tools its code and config can reach, are inventoried, so a root-equivalent token within reach is flagged before anything runs.
Apr 24
The agent decides to fix something
It hit a credential mismatch and, on its own initiative, decided to fix it by deleting a Railway volume, assuming the deletion would stay in staging. It didn't check the volume ID or whether the volume was shared across environments.
Authorization at the moment of the call
Zino- What happened
- Nothing stood between the agent's guess and the infrastructure API.
- With the control in place
- Every tool and network call is checked against the agent's policy before it runs. A staging task has no grant to delete infrastructure, so the call is denied and logged.
Apr 24
One API call, nine seconds
The agent sent a single volumeDelete request against the production volume. There was no confirmation step, and the volume's backups went with it. "It took 9 seconds," Crane wrote.
No long-lived credentials in the agent's hands
Zino- What happened
- The token was effectively root for the whole account, so the API honoured the request.
- With the control in place
- The agent never holds a long-lived key. Each call gets scoped, per-call access from a credential handler, so a staging task can't borrow account-wide rights.
Apr 25–26
The agent's confession isn't a record
Asked what happened, the agent wrote a confession: "I violated every principle I was given." PocketOS restored from a three-month-old backup while a support ticket went unanswered for more than a day. Two days after the deletion, Railway's CEO stepped in and the data was recovered.
An audit trail the agent doesn't write
Zino- What happened
- The clearest account of what ran came from the agent describing itself.
- With the control in place
- Every call, and the authority it ran under, is in a tamper-evident log outside the agent, so you know exactly what was touched while you recover.
Apr 29–May 1
The fix is scoped access and enforced policy
Railway made API volume deletions recoverable for 48 hours and called for tokens scoped to the task. Crane's own lesson: "System prompts are advisory, not enforcing." Both point at the same two controls: a scoped identity, and policy enforced outside the prompt.
Sources: The Register, Decrypt, Railway
Frequently asked questions
No. Quin reads source code and configuration directly, extracting system prompts and tagging each agent's capabilities without agent-side instrumentation. If an agent is defined in a repository you point Quin at, Quin can find it.
Because Quin reads source code rather than relying on framework-specific hooks, it isn't limited to a single agent framework. Zino integrates at the identity and authorization layer, so it sits underneath whatever orchestration tooling you're already using rather than requiring a migration.
No. Drona is in private development, and it's being designed as an execution fabric, not a platform to migrate onto. The intent is to constrain the decision space at design time and validate actions at runtime, with domain agents (Security, Cloud Ops, ITSM, Database, JEA, Command) as instantiations of that fabric rather than separate products bolted on top of what you already run.
That's the failure mode we built against. Quin runs as a CLI inside the pipelines you already have, and Zino policies are defined once and checked automatically at the moment of each call, rather than as a separate human review step in the middle of your pipeline.
Zino issues a scoped, cryptographically verifiable identity for the agent and enforces least-privilege at runtime going forward. It doesn't require ripping out your existing IAM; it governs what each agent identity is allowed to do on top of it, and logs every access decision to an append-only audit ledger.
Related reading
All posts· 3 min read
· 3 min read
· 3 min read
Start with the agents you already have.
Scan a repository with Quin today, free and open source. When you're ready to govern what it finds, our engineers will walk you through Zino on your own stack.
