Boltastic

Action at the speed of thought

Comment 1

Boltastic is a multi-agent AI observability and GTM copilot that connects Engineering, Marketing, and Sales through a unified intelligence layer.

It combines Elastic observability, Salesforce CRM, Jira, Slack, Git workflows, MCP tooling, and multi-agent orchestration into a single conversational experience.

Instead of siloed dashboards and reactive workflows, Boltastic detects issues, surfaces growth opportunities, and coordinates action across systems — with humans always in the loop.

🚀 The Problem

Revenue teams operate across disconnected tools:

  • Marketing sees metrics but struggles to scale success.
  • Sales notices deal risk too late.
  • Engineering receives vague issue reports without telemetry context.

Observability tools show signals. CRMs track activity. Dashboards visualize trends.

But nothing connects them into coordinated, intelligent execution.

Boltastic fills that gap.

🧠 System Architecture

Architecture Diagram

Boltastic is built as a distributed multi-agent system.

🔹 Supervisor Layer – Google ADK

Supervisor Agent

The Supervisor Agent is built using Google ADK.

It:

  • Serves as the single conversational interface.
  • Interprets user intent.
  • Dynamically routes queries to the appropriate domain agent.
  • Orchestrates multi-agent collaboration when needed.

The Supervisor does not directly query tools — it coordinates agents.

🔹 Domain Agents – Elastic Agent Builder

The domain agents are built using Elastic Agent Builder:

  • Core Tech Agent – Engineering & observability
  • Nova Agent – Marketing intelligence
  • Orbit Agent – Sales pipeline & CRM

These agents live within Elastic’s ecosystem and are optimized to:

  • Query Elastic indices using ES|QL
  • Leverage anomaly detection jobs
  • Trigger connectors and workflows
  • Enforce strict schema constraints

🔹 A2A Protocol Communication

The Supervisor (Google ADK) communicates with Elastic agents via the A2A (Agent-to-Agent) protocol.

This allows:

  • Decoupled orchestration
  • Cross-platform agent coordination
  • Clean separation between routing logic and domain execution
  • Scalable multi-agent expansion

This architecture keeps Boltastic modular and extensible.

🔹 MCP Integrations

Boltastic also connects to MCP tools for advanced capabilities.

The most powerful example is the custom ui_analysis_tool, which:

  • Launches Playwright against the live frontend.
  • Captures UI screenshots.
  • Sends visual + contextual data to Gemini.
  • Returns structured analysis back to the calling agent.

MCP connections allow agents to:

  • Interact with external systems.
  • Execute controlled workflows.
  • Extend beyond Elastic-native capabilities.

🔧 Core Tech Agent – Observability to Auto-Remediation

Core Agent Flow

The Core Tech Agent connects telemetry, UI context, and the codebase.

It can:

  • Analyze logs and performance metrics in Elastic.
  • Detect anomalies and error spikes.
  • Visually inspect the live application using MCP + Playwright.
  • Correlate telemetry with UI state using multimodal reasoning.

And when an issue is well-defined, it can:

  1. Navigate the project directory.
  2. Identify relevant source files.
  3. Generate and apply a fix.
  4. Create a new Git branch.
  5. Commit and push changes.
  6. Open a pull request with structured reasoning.

All PRs require human review before merge.

This closes the loop from detection → diagnosis → code-level resolution.

📈 Nova Agent – Marketing Intelligence

Nova Agent Flow

Nova continuously monitors campaign performance using:

  • Elastic indices
  • ML anomaly detection jobs

It:

  • Identifies negative anomalies (CTR drops, conversion decline).
  • Identifies positive anomalies (viral spikes, outperforming campaigns).
  • Converts growth spikes into structured scaling opportunities.
  • Recommends budget shifts or experiments.
  • Triggers Jira tasks or Slack notifications through Elastic connectors and workflows (after approval).

Nova treats positive anomalies as growth signals — not just noise.

🎯 Orbit Agent – Sales Pipeline Intelligence

Orbit Agent Flow

Orbit bridges Elastic and Salesforce.

It:

  • Flags stalled or high-risk deals.
  • Analyzes inactivity patterns and deal size.
  • Queries official Salesforce playbooks.
  • Proposes structured interventions.
  • Creates deal_interventions__c records after confirmation.

Through Elastic workflows and Salesforce integrations, it turns CRM data into proactive deal recovery.

🔄 Workflows & Connectors

Boltastic leverages Elastic connectors and workflows to move from insight to action.

Examples include:

  • Anomaly detected → Slack notification
  • Growth spike → Jira scaling task
  • High-risk deal → Salesforce intervention record
  • Confirmed engineering issue → PR + Jira ticket

Workflows are defined declaratively and executed safely, ensuring:

  • Structured payloads
  • Schema enforcement
  • Human approval before mutating actions

This allows Boltastic to not just analyze data — but orchestrate execution.

⚙️ Tech Stack Overview

Supervisor Layer

  • Google ADK

Domain Agents

  • Elastic Agent Builder

Agent Communication

  • A2A protocol

Frontend

  • React + Vite Supervisor UI

Observability

  • Elastic dashboards
  • ES|QL queries
  • ML anomaly detection jobs

Integrations

  • Salesforce (SOQL + record creation)
  • Jira & Slack connectors
  • Git branch + PR automation

MCP Server

  • Python + FastMCP
  • Playwright
  • Gemini multimodal reasoning

🏆 What Makes Boltastic Different

  • Cross-platform multi-agent architecture (Google ADK + Elastic Agents)
  • A2A-based orchestration
  • Observability + CRM + Git in one loop
  • Autonomous PR generation from production telemetry
  • Positive anomaly detection for growth acceleration
  • Human-in-the-loop governance baked into every mutation

🌟 Vision

Boltastic becomes the intelligent coordination layer between:

  • Observability systems
  • CRM platforms
  • Marketing analytics
  • Engineering codebases

Not just reporting signals — but orchestrating decisions and executing improvements across the revenue organization.

Built With

Share this project:

Updates