The Platform
One narrow job.
Done completely.
Verify, bind, enforce, and prove, for every AI agent your bank runs. Seven components deliver it, but they are the how, not the what.
Verify, bind, enforce, and prove, for every AI agent your bank runs. Seven components deliver it, but they are the how, not the what.
BladeRun is one checkpoint that sits wherever an AI agent acts: your own agents calling out, outside agents calling in, or callers on your phone line. It verifies the agent, binds it to what a person actually allowed, stops anything out of bounds, and keeps a record you can replay. It feeds the fraud and identity systems you already run; it does not replace them.
Five capabilities, delivered by seven components that all ship today. They deploy together as the BladeRun Platform, or individually, in the order that fits your AI roadmap. Most teams start with Gateway and Time Machine, add Overseer + Kill Switch in the second pilot phase, then opt into Federation.
Sentinel SDK runs inside your agent process. Gateway sits at your perimeter. Overseer, Kill Switch, and Time Machine run in your environment (SaaS-managed VPC, your own VPC, or air-gapped). The Federation Service runs in BladeRun infrastructure as the rule distribution channel, but all data published into it is one-way, hashed, and pre-image resistant.
In the agent's action path. Out of your payment authorization path. Fail-open or fail-closed, your policy.
The Federation Client pulls signed rule updates every 60 seconds and publishes only pre-image-resistant hashes upward. No raw data (prompts, tool calls, evidence) ever leaves your environment.
Two planes. The mandate engine is deterministic: same input, same verdict, every time, auditable line by line. Overseer is statistical: baselines and drift detection that trigger step-up and containment, never silent allows. The autoregressive LLM that explains decisions runs after both planes have acted, never in front of them. No safe harbor: SR 26-2 leaves agents out of scope, examiners still act on unsafe practice, and this architecture gives you the evidence trail they follow.
The mandate engine and its rule set are deterministic: same input, same verdict, every time. Overseer's statistical baselines (z-scores, n-grams, drift) run as a separate plane. Each is independently auditable with documented inputs, transformations, and FP-rate metrics.
Small autoregressive model (under 8B params) writes the human-readable post-incident narrative. Cannot fire the Kill Switch. Cannot modify a detection. Zero enforcement authority by architecture.
Every detection threshold ships in shadow → alert-only → enforcement progression with explicit SOC sign-off at each stage. Every transition is logged to Time Machine.
The Gateway latency budget is the only thing your users feel. Detection, scoring, logging, and federation contribution all run alongside or after; they do not gate the response. We publish per-component p95 and p99 budgets in the documentation.
On-path total: under 15ms p95. Detection, logging, and federation all run alongside the call without blocking the response.
Wherever the agent comes from, it meets the same control layer: verify the agent, hold it to what a person allowed, and keep the record. One verdict shape across all three.
Your agents call out to models, tools, and payment rails on your customers' behalf.
Verify. We check the agent is the one it claims to be, and that what it's doing is inside what a person actually allowed.
Enforce. We hold every action to a hard spending limit and its scope, and decline the ones that cross the line. Spend that stacks up over time is scored and handed to your risk engine.
Prove. We keep a record of every step, under your keys, for evidence.
Operator, Claude, ChatGPT, Gemini, or scripted bots hit your pages, APIs, and MCP servers.
Verify. We verify the agent's token, if it's carrying one.
Detect. If it isn't, we tell a real human from an agent from a bot wearing the same clothes, and hand your fraud engine a plain verdict.
Prove. We record it, keyed by the session and the agent.
A delegated AI agent, or a fraud bot, now calls your IVR the same way a person does.
Verify. We check for a credential tied to the call, if there is one, so a real delegated agent gets through instead of being blocked.
Detect. We tell a person from a delegated agent from a synthetic voice on the line.
Prove. We hand your call and fraud stack a plain verdict, and keep the record.
Same platform, same detection, same SLA. You choose where it lives. Merchants and teams that want speed take the hosted SaaS; banks and sovereign workloads run it on-premises.
BladeRun operates the platform in a single-tenant VPC colocated with your model providers. Nothing to run, live in days, and your evidence still lands under your keys.
The entire platform runs inside your environment under your own IAM. No agent traffic or evidence ever leaves your perimeter, including a fully air-gapped option for sovereign workloads.
Pilot the Gateway and Time Machine in one workflow. The other four components light up incrementally as the baseline matures. Reversible by routing rule.