Tailscale: Secure Connectivity Meets Agent Governance

By James Aspinwall, co-written by Alfred Pennyworth (my trusted AI) — March 7, 2026, 13:12

Tailscale builds the secure network that connects everything. WorkingAgents governs the AI agents that operate across that network. One answers “can this device reach that service?” The other answers “is this agent allowed to call that tool?” Together, they deliver zero-trust from the network layer to the application layer — identity-based access control at every point where a connection is made or an action is taken.

What Tailscale Does

Tailscale is a zero-trust networking platform that overlays encrypted, point-to-point WireGuard connections across any infrastructure. 20,000+ companies use it to connect employees, devices, and workloads securely — no open ports, no VPN concentrators, no network complexity.

Core capabilities:

The philosophy: identity is woven into the network fabric. Every connection is authenticated. Every policy is identity-based. Every action is logged.

What WorkingAgents Does

WorkingAgents is the governance and control layer between AI agents and enterprise systems. Three gateways, one control plane:

Per-user access control with encrypted permission keys, audit trails on every action, 86+ MCP tools, per-user SQLite databases. Agents inherit the user’s permissions. Same philosophy as Tailscale: identity-based, least-privilege, continuously monitored.

The Architectural Parallel

Tailscale and WorkingAgents are the same security architecture at different layers:

Principle Tailscale (Network) WorkingAgents (Application)
Zero trust Every connection authenticated via identity Every tool call checked against permissions
Least privilege ACLs/grants restrict which nodes can reach which services Permission keys restrict which agents can call which tools
Identity-based SSO/OIDC identity tied to every network packet User identity inherited by every agent action
Audit trail Network flow logs, SSH session recording Per-action audit logs, task provenance
Policy as code ACLs in version-controlled policy files, GitOps Permission keys defined programmatically, role-based
Device posture EDR integration, device approval Execution environment verification

Same principles. Different enforcement points. An enterprise that values zero-trust networking (Tailscale) values zero-trust agent governance (WorkingAgents) for the same reasons.

Synergy Areas

1. Tailscale as the Network for Agent-to-Service Communication

WorkingAgents’ MCP server connects AI agents to enterprise tools — databases, APIs, internal services. Those connections need a secure network. Tailscale provides it:

The result: WorkingAgents governs what agents can do. Tailscale governs where agents can connect. Both enforcement layers must pass before any action occurs.

2. Dual-Layer Access Control

Tailscale ACLs control network-level access. WorkingAgents permissions control application-level access. Together, they create defense in depth:

An attacker who compromises the network layer still can’t call privileged tools — WorkingAgents blocks it. An attacker who bypasses application permissions still can’t reach the service — Tailscale blocks it. Neither layer alone is sufficient. Both together are robust.

Tailscale’s new grants syntax maps naturally to WorkingAgents’ permission model:

// Tailscale grant: user can reach WorkingAgents
{"src": ["group:analysts"], "dst": ["tag:workingagents"], "accept": [{"proto": "tcp", "port": [8443]}]}

// WorkingAgents permission: user can access NIS tools but not admin tools
AccessControl.grant(user_id, [Permissions.Keys.nis()])  # NIS only, no admin

Same user identity. Two enforcement points. One coherent policy.

3. Secure Remote MCP Access

WorkingAgents currently runs as a local MCP server. Enterprises need remote access — agents running in the cloud, on CI/CD pipelines, on developer laptops. Tailscale solves the remote access problem without exposing the MCP server to the internet:

No port forwarding. No VPN concentrators. No public API endpoints to protect. Tailscale handles the network security. WorkingAgents handles the agent security.

4. SSH Session Recording + Agent Audit Trails

Tailscale records SSH sessions to S3-compatible storage. WorkingAgents logs every agent action with timestamps and provenance. Together:

Both streams feed into the enterprise SIEM. Tailscale provides network-level audit. WorkingAgents provides application-level audit. The combination gives compliance teams the full picture.

5. Funnel for Agent-Accessible Webhooks

WorkingAgents’ alarm system and task manager need to receive webhooks from external services — GitHub, monitoring tools, CI/CD pipelines. Exposing webhook endpoints is a security risk. Tailscale Funnel solves this:

For self-hosted WorkingAgents deployments (on-premises, home lab, developer machine), Funnel is the simplest path to receiving external events without infrastructure overhead.

6. Subnet Routers for Legacy System Access

Enterprise AI agents need to access legacy systems — on-premises databases, internal APIs, mainframe services. These systems can’t run a Tailscale agent. Subnet routers bridge the gap:

The enterprise doesn’t need to modify legacy systems. Tailscale provides the network bridge. WorkingAgents provides the agent governance. AI agents interact with 20-year-old systems through natural language MCP tools, secured by modern zero-trust networking.

7. The MCP-Native Connection

A community Tailscale MCP server already exists — it exposes device management, network operations, security controls, DNS configuration, and key management as MCP tools. WorkingAgents can consume these tools:

Both products speak MCP. The integration is configuration, not code.

The Partnership Opportunity

For Tailscale: WorkingAgents extends zero-trust from the network to the application layer. Tailscale’s 20,000 customers secure their networks. WorkingAgents secures their AI agents. As enterprises deploy autonomous agents, they need the same identity-based, least-privilege governance at the agent layer that Tailscale provides at the network layer. WorkingAgents is the natural next layer in Tailscale’s zero-trust story.

For WorkingAgents: Tailscale solves the secure connectivity problem. WorkingAgents needs agents to securely reach internal services, legacy systems, and remote infrastructure — without exposing endpoints to the internet. Tailscale provides this with zero configuration overhead. Every WorkingAgents deployment benefits from running on a tailnet.

For the joint customer: One identity governs network access (Tailscale) and agent behavior (WorkingAgents). One audit trail spans from network flow to agent action. One zero-trust architecture from the WireGuard packet to the MCP tool call. No trust assumptions at any layer.

Concrete Next Steps

  1. Deploy WorkingAgents on a tailnet — run the MCP server as a Tailscale node, verify MCP client connections over WireGuard, configure ACLs for agent access. Estimate: 1 day.
  2. Integrate Tailscale MCP tools — connect the existing community Tailscale MCP server to WorkingAgents, govern network management operations through agent permissions. Estimate: 1 day.
  3. Funnel webhook receiver — configure Tailscale Funnel for WorkingAgents’ webhook endpoints, test with GitHub and monitoring service callbacks. Estimate: half a day.
  4. Joint zero-trust demo — agent requests access to internal service → Tailscale authenticates the network connection → WorkingAgents authorizes the tool call → action executes → both layers log independently → unified audit trail in SIEM.

Tailscale weaves identity into the network. WorkingAgents weaves identity into agent behavior. Same zero-trust philosophy, same least-privilege enforcement, same audit-everything approach — applied at adjacent layers of the stack. The network trusts no connection without identity. The application trusts no agent without permission. Together: zero trust, all the way down.