Shield, Protect, product 2 of 8
The problem. Agents can pretend to be someone else, sign in with a customer’s password, or scrape data you meant to keep private.
What changes. Every agent proves who it is, and your policy decides what it may see and do, wherever it arrives.
Working in lab runs end to end in our lab todayIn lab on our build plan: designed, being built or tested in our lab; not available yetWhat works today
Pick a request and follow it through the checks every agent request passes.
Illustrative simulation of requests from AI agents passing through the TrueSession gateway. Choose a request below to see how each check is answered.
Incoming request
Shopping assistant
Signed request
“Buy the trail running shoes in size 10.”
Your TrueSession gateway
4 checksVerified agent
Who is really asking?
Your policy
What may this agent do here?
Live truth
Is it accurate right now?
Customer approval
Did the customer agree to exactly this?
Your business
Nothing reaches your systems until the checks pass.
Every request is signed by the agent and checked against its published key, for freshness and for replays. Knowing who the agent is unlocks nothing on its own.
Decide what stays public, what only verified agents may see, and which actions need the customer. A script only points the way; enforcement runs at your server, backend and CDN, where every request arrives.
The customer grants an agent specific access, a limit and an expiry, inside what your policy allows. The agent never holds the customer’s password, and one revoke ends its access.
Some agents ignore your agent address and use your website like a person. TrueSession spots them by how they behave, flags the whole visit on your server and points them to your agent address. Your policy decides what the flag changes, and behaviour alone never blocks anyone or unlocks anything.
Your policy decides what every agent may do at your store. Change a rule and watch the requests coming in. Some protections are always on and can’t be switched off.
Policy editor:In lab
Requests coming in
Decided by your policy, in code.
Signatures, short time windows and one-time checks refuse them.
Agents get a scoped, revocable grant instead, within your policy.
Text on pages is treated as data. A misled agent still can’t act outside its access.
Stock, partner rates and account data need a valid, short-lived access grant.
Payment happens on your provider’s page. Agents never see card numbers.
Agents get answers from your systems and the facts you reviewed.
No product can detect every agent. An agent working inside a customer’s own browser can look exactly like the customer. That is why the actions that matter always need the customer’s approval.
Each feature with its status today.
No script can do that alone, because many agents never run JavaScript. TrueSession uses the script for discovery and handoff, and enforces at your server or CDN edge, where every request arrives. Requests without a valid grant cannot reach protected data or complete consequential actions.
No one can identify every visitor. An agent working inside a customer’s own browser session can look identical to that customer. When behaviour suggests an agent, TrueSession flags the visit on your server and points it to your agent address, and your policy can slow it down or ask for a check. Behaviour alone never blocks anyone or unlocks anything. And even an undetected agent cannot complete a purchase, transfer or account change without the customer’s exact approval.
Bot management decides which traffic reaches your site. MCP gateways turn APIs into tools for a company’s own agents. TrueSession is for other people’s agents acting for your customers: it adds your policy per action, the customer’s delegation and exact approval, payment confirmation, signed answers, conversations and proof, across every agent protocol, and it works with the CDN you already have.
We are designing pilots with merchants, marketplaces and banks.