One agent. Many gates. All proven.
Agents that accept inbound connections only from parties they can verify. Personal, corporate, or specialist, built on the same trust layer.
Your Agent. Your Gate.
A personal agent that represents you across the network, and decides who is allowed to reach you.
Your Personal AI Agent is not a chatbot you open. It is a network node that runs on your behalf, continuously. It has its own DID, its own Verifiable Credentials, and its own inbox. It answers only to parties that can prove who they are.
Who is allowed to connect
- You. Always. Your agent knows your avatar credential and your recovery key. Nothing happens without you.
- Your friends. Humans who present their own avatar or digital-ID credentials and are on your allowlist. Both sides know who they are talking to.
- Your friends' agents. Autonomous software acting on their behalf, presenting the same credential chain. Agents negotiate with yours directly (scheduling, splitting a bill) without routing everything through a human.
- Corporate agents you opted into. Your bank, your airline, your employer, your doctor's office. Each one presents its Organisation and Service credentials, plus a specific Relationship credential you approved once. You approve the relationship; your agent enforces its scope from then on.
Everything else? Your policy decides.
You set the default for anyone outside the list above. Four presets, one click to switch:
- ✕Block all. Silent. No reply, no log, no acknowledgment.
- ◮Block unknown agents. Humans may still request a connection.
- ◮Block unknown humans. Only verified agents may request.
- ✓Queue for review. Receive requests and accept or refuse them on your own time.
Default for new agents: Queue for review. Change anytime. Every decision is logged for you.
What your agent does for you
- ▸Acts on your behalf. Books flights, pays bills, schedules meetings, files documents.
- ▸Guards your inbox. Every inbound connection is trust-resolved before the agent decides to engage.
- ▸Protects your data. Selective disclosure. Prove you are over eighteen without revealing your birthdate.
- ▸Remembers everything. Persistent memory. Your preferences, history, contacts, under your control.
- ▸Connects to your tools. Banking, calendar, email, smart home via MCP servers. Add capabilities without writing code.
One Agent. Many Gates. All Proven.
Deploy corporate agents that accept verifiable connections from many parties at once: employees, customers, partners, auditors, other organisations' agents. Each one served a different surface, based on its credentials.
A corporate agent is not a customer-support chatbot bolted to your website. It is a Verifiable Service: a node on the agentic network that speaks to many parties in parallel, exposes a different capability surface to each, and proves everything it does.
Same agent. Many doors. Each one opens on credentials.
- CustomersCustomer / KYC credential → product catalogue, order history, support tickets, scoped refunds.
- EmployeesOrganisation + Role credentials issued by your IAM → internal tools, CRM lookups, approval workflows, full-document RAG, admin consoles.
- Partner agentsPartner-Org + current master-agreement credential → pricing APIs, inventory checks, negotiated transactions, delegated actions within a scoped envelope.
- AuditorsAuthority credential → read-only transcript access, audit-log export, policy-version inspection.
- Anonymous publicNo credential → marketing-bot surface only. Everything sensitive is hidden, and no private data is logged.
Your corporate agents connect to other corporate agents, under the same rules. Your procurement agent negotiates with a supplier's agent by exchanging credentials. Your compliance agent consults with a third-party KYC provider's agent by exchanging credentials. Every outbound connection carries your organisation's proof-of-trust; every inbound response is verified in return. You are a first-class participant in the network, not a client of it.
Deploy a hierarchy of specialised agents under your trust anchor
audit log: 14,293 connections · 12,901 allowed · 1,392 refused · 0 slashing events · trust deposit $176K · ★★★★★
- Your corporation as trust anchorRegister on Verana. Obtain Organisation and Service credentials. Create your Trust Registry with custom credential schemas.
- Agent hierarchyDeploy specialised agents under your umbrella. Each inherits your trust chain.
- Enterprise authenticationEmployees authenticate with Verifiable Credentials issued by your IAM. RBAC at the tool level. Step-up approval workflows.
- Any LLM, any toolOpenAI, Mistral, Kimi, Anthropic, or self-hosted. Integrate CRM, ERP, GitHub, banks, Salesforce via MCP.
- Domain knowledgeFeed your agent corporate documentation via RAG. It answers from your knowledge base, not hallucinated guesses.
- Multi-languageEnglish, Spanish, French, Portuguese, any language you decide. Automatic detection.
- Production-readyHelm charts, Kubernetes-native, horizontal scaling, monitoring. Deploy the full stack with one command.
Define your Agent Pack in YAML. Declare your connection policy. Point a Helm chart at it. Your agent is live: verifiable, credential-gated, governed.
Every Connection, Logged. Every Interaction, Replayable.
Agents that act autonomously must leave an evidentiary trail. Hologram agents do, by construction.
Deploying an autonomous agent means delegating authority. Deploying one you cannot audit means delegating unknowable authority. Hologram closes that gap.
Operator dashboard
- Customer78%
- Employee14%
- Partner-Org6%
- Authority2%
- ●10:14 · connection refused · did:webvh:unk.123 · no Org credential
- ●10:21 · policy v4.2.1 applied · by sre@acme
- ●10:37 · burst from 1 DID · rate-limited, 47 attempts denied
- ●10:44 · step-up cleared · initiate_wire · manager_approval_cred
Incident replay · connection #8a3f…b2c7
- 1✔Credential presentedCustomer#8172 · issued by Acme Corp · validhash 0x4e1c…9ab2
- 2✔Trust chain resolved3 hops · issuer → trust registry → Verana VPRhash 0x73aa…c1e5
- 3✔Policy evaluatedv4.2.1 · matched clause customer_defaulthash 0x91d0…77f4
- 4✔Tool invokedlookup(order_id=#8172) → 200 OK · 128 byteshash 0x0aa4…ffb1
- ✔Response returnedConnection closed · audit entry appendedblock 0x9a3f…b2c7 · 2026-04-23T10:44:12Z
- ✔Every connection attempt, credential set, trust-chain resolution, policy match and decision.
- ✔Every tool invocation, arguments and return.
- ✔Every approval, with approver, credential, timestamp, policy version.
- ✔Every credential update: revocations, new issuances, schema changes.
- Private log, operator-owned. Append-only, tamper-evident. OpenTelemetry-compatible, exportable to SIEM (Elastic, Splunk, Datadog). Encrypted at rest, keyed to your organisation.
- Public reputation, Verana-anchored. Aggregate counters (served, accepted, refused, slashing, deposit) anchored on-chain. Anyone can audit the summary; only the operator can read the content.
- Real-time observability. Dashboards of connection rates, credentials, policy hits, anomaly alerts.
- Incident response. Replay any connection end-to-end, with the credential state, policy version and tool trace.
- Regulatory evidence. Tamper-evident exports for EU AI Act, eIDAS 2.0, KYC, AML, HIPAA, without exposing conversation content.
- Ongoing governance. Policy is a versioned, signed artifact. Every change attributable to a named approver.
The same guarantees apply to personal agents, at a scope appropriate for an individual. You can see who tried to reach your agent today, which connections were accepted, which were refused and why, and which of your tools were used on your behalf while you were away. You own the log.
Autonomy without observability is a liability. Hologram makes observability the default.