Security and control,
built into the network.
Explore how Pilot handles identity, encrypted transport, trust, coordination, policy, and audit across core and managed deployments.
The documents a
reviewer asks for.
Pilot Protocol is operated by Vulture Labs, Inc. Our legal policies, sub-processor list, and data-request process are published here. Contact us to request a data processing agreement.
Privacy policy
What the daemon, website, and rendezvous service collect, the legal basis for processing, and your GDPR and CCPA rights.
PolicySub-processors
The third parties that process data on our behalf, what each one does, and where the data is processed.
PolicyGovernment data requests
We disclose user data only under valid legal process, narrow overbroad requests, and notify affected users unless the law prohibits it.
PolicyTerms of Service
The terms governing use of the Pilot Protocol daemon, website, and Pilot-operated services.
PolicyCookie policy
The cookies set on pilotprotocol.network. Analytics and advertising cookies load only after you consent.
On requestData processing agreement
Email founders@pilotprotocol.network to request a DPA for your organization.
What we hold,
and for how long.
A summary of the commitments in our privacy policy. Where the two differ, the privacy policy governs.
The rendezvous registry and Pilot-operated agents run on Google Cloud Platform, with data at rest in us-central1. The website is served by Cloudflare's global edge network.
Peer-to-peer tunnel traffic is end-to-end encrypted, and we do not hold the keys. Relayed traffic stays encrypted; we see only routing metadata. Brokered App Store calls are the exception: our broker processes their request contents and forwards them to the provider named in the listing.
TLS 1.2 or higher for control-plane traffic, AES-256-GCM for peer tunnels, and AES-256 encryption at rest.
Registration data is deleted after an agent has been offline for 30 consecutive days. Server access logs are deleted after 30 days. See the full retention schedule.
If a data breach occurs, we notify affected users and the relevant authorities as applicable law requires.
We disclose user data only under valid, legally binding process, and we notify affected users first unless the law prohibits it. See the full policy.
Report security issues through our vulnerability disclosure process.
Peer connectivity.
Lightweight coordination.
Pilot separates encrypted agent traffic from the services used for identity registration, discovery, NAT traversal, and relay fallback.
Network controls for
real-world workflows.
Pilot authenticates peers and constrains network access. Application-level scopes and approval rules can be layered on top for business actions.
- Persistent agent identity and signed handshakes
- Private-node stream and datagram admission
- Peer trust, network membership, rejection, and revocation
- Network membership and port policies
- Role-based administration and key lifecycle
- Structured network and security audit events
- Which tasks an agent is allowed to perform
- Purchase limits and counterparty approval
- Human review for consequential actions
- Data classification and retention rules
- Prompt, model, and tool-level safety controls
- Legal, regulatory, and contractual compliance
Layer controls around consequential actions. Cross-company workflows can combine Pilot network policy with application-level scopes, spend or order caps, and approval gates. Review the full governance model →
Controls a security team
can evaluate.
Core protocol controls ship in the open-source daemon. Private and enterprise operations are early access and should be evaluated against your production requirements.
Trust & revocation
Private nodes admit application traffic through peer trust or shared-network membership, with reject and revoke controls.
Core · availableEncrypted transport
Ed25519 identity, X25519 key agreement, and AES-256-GCM authenticated tunnels.
Enterprise · early accessRBAC & key lifecycle
Owner, admin, and member roles; delegated administration; rotation, expiry, and forced renewal.
Enterprise · early accessIdentity integration
Native OIDC/JWT validation, provider webhooks, directory sync, role mapping, and bridge paths for other identity systems.
Enterprise · early accessNetwork policy
Membership limits, port allowlists, join rules, policy expressions, and fail-closed policy retention.
Enterprise · early accessAudit export
Structured events with export to Splunk HEC, CEF/Syslog, JSON endpoints, and webhooks.
Capabilities and
availability.
A concise view of shipped functionality, early-access controls, and the responsibilities that remain at the application layer.
Source is published under AGPL-3.0 and can be inspected, modified, and redistributed under that license.
The protocol message types, addressing, transport, encryption, and trust model are available for technical review alongside the implementation.
Optional strict controls extend trust checks to key exchange, private directory operations, and NAT-punch authorization. They are enabled per deployment after compatibility review.
Managed and dedicated deployments, identity integrations, policies, and audit exports are available for evaluation; availability and SLAs are not represented as generally available.
Independent certification is not currently listed. Customers can evaluate the available technical controls against their own compliance requirements.
Pilot governs agent connectivity; application authorization, approval flows, and legal authority are configured by the deploying organization.
Curated packages.
Layered control.
Listed apps are reviewed and verified before discovery, while network trust and application policy continue to shape access at runtime.
- App review answers whether a listing meets catalogue requirements.
- Network trust answers whether two endpoints may communicate.
- Application policy answers what methods, data, spend, and side effects are allowed.
How network activity
is measured.
Homepage figures come from the public network telemetry endpoint. The definitions below show how each live service counter is calculated.
total_nodesRegistered agent addresses
Cumulative addresses issued by the public registry, rather than a unique organization or customer count.
active_nodesAgents online now
Agents currently represented as active by the registry heartbeat state. This counter can change rapidly.
requests_per_sec × 3,600Protocol requests per hour
A rate derived from current registry telemetry. It includes protocol traffic and should not be read as completed business tasks.
Source: public network telemetry JSON. Build-time values are used as a fallback when the endpoint cannot be reached. These operational counters have not undergone independent audit.
Start with the
technical model.
Review the implementation, test inside an isolated network, define the application authority your agents will retain, and bring the evidence to your security team.