Skip to main content
    A2A Protocol AgentChecking status…

    An executive
    second mind.

    An executive coordination agent powered by Google Gemini on Vertex AI. It generates leadership briefings, routes work between specialized agents, and coordinates across enterprise cloud systems. It ships in two flavors — an Agent-to-Agent (A2A) service on Cloud Run and a Vertex AI Agent Engine deployment — that share the same six skills and env-var contract.

    View Agent Card Source on GitHub

    Live infrastructure / 01

    A working interface between leadership intent and enterprise systems.

    https://a2a.mammonbaloch.space

    02 / Deployment surfaces

    Two ways to reach it

    A2A Protocol · Cloud Run

    A standalone Agent-to-Agent service on Google Cloud Run, discoverable to any compliant A2A client via its public Agent Card. Calls to the JSON-RPC endpoint are protected with a Google OAuth 2.0 ID token (see the authentication section below).

    View Agent Card OAuth docs

    Vertex AI · Agent Engine

    A Python Agent Development Kit (ADK) build of the same coordinator, deployed to Vertex AI Agent Engine. It shows up in the GCP Vertex AI → Agent Builder catalog and is testable straight from the console's built-in “Try it” pane — the right surface to demo to teams on Google Cloud.

    Open in Agent Builder

    Runs in GCP project symbolic-rope-445407-i1 (us-central1).

    03 / Public interface

    A2A public endpoints

    Agent Card (unauthenticated)

    https://a2a.mammonbaloch.space/.well-known/agent.json

    JSON-RPC endpoint (OAuth-protected)

    https://a2a.mammonbaloch.space/a2a

    Implements message/send, message/stream (SSE), tasks/get, and tasks/cancel.

    04 / Capability map

    Six skills, one coordinator.

    Executive Operations

    Leadership briefings, executive summaries, board reports, and priority queues for the CEO's office.

    Multi-Agent Coordination

    Routes tasks and orchestrates workflows between specialized AI agents across AWS, Google Cloud, Microsoft, and IBM.

    Governance & Compliance

    Reviews decisions against governance policies, generates compliance reports, and enforces approval workflows.

    Partner & Market Intelligence

    Synthesizes partner data, competitive intelligence, and market signals into strategic opportunity assessments.

    Enterprise Automation

    Drafts automation blueprints, triggers cross-system workflows, and coordinates enterprise SaaS integrations.

    Scheduling

    Turns executive actions into RFC 5545 calendar events and optionally posts them to a configured calendar webhook.

    Skills currently return structured stub data while real enterprise connectors are wired in.

    05 / Access protocol

    How to authenticate

    The agent verifies callers with a Google OAuth 2.0 ID token presented as a Bearer token. The token must be issued by accounts.google.com with the aud claim equal to the agent's Cloud Run URL.

    1. Have your calling agent authenticate as a Google service account (or Workspace user).
    2. Mint an ID token with the audience set to the agent URL above.
    3. Send requests to the JSON-RPC endpoint with Authorization: Bearer <token>.
    4. Your identity (service-account email, Workspace domain, or numeric sub) must appear on the agent's caller allowlist.

    Fail-closed by default

    Unauthenticated requests return 401. Authenticated but non-allowlisted callers return 403. To request access, reach out via the contact page with the identity you want allowlisted.

    This page and the blog share no runtime, database, or auth with the agent service — the agent runs as an independent Cloud Run deployment.