OpsPulse - Alexa+ Autonomous MCP Agent

Simulated Alexa+ developer copilot leveraging Streamable HTTP MCP tools and AWS Bedrock for voice-driven devops triage and repository automation.

Comment

Inspiration

Software engineers frequently juggle terminals, GitHub dashboards, CI/CD pipelines, and logs. This constant context-switching destroys focus and productivity. We wanted to see what happens when developer operations meet the ambient intelligence of Alexa+: what if triaging repository status, running security audits, and drafting release notes was as simple as having a casual voice conversation in the workspace or living room?

What it does

OpsPulse is a simulated Alexa+ autonomous developer copilot built on the open Model Context Protocol (MCP Spec 2025-11-25) and AWS Bedrock.

  • Voice-Driven Dev Triage: Responds to natural voice/text commands (e.g., "Alexa, run a security audit", "Alexa, scan repo health", "Alexa, draft release notes").
  • Self-Hosted MCP Engine: Conforms to the Streamable HTTP MCP standard. It dynamically advertises tools (tools/list), routes intent, and runs live diagnostics without manual pipeline triggering.
  • Ambient Feedback & Speech: Features an interactive glowing Alexa ambient orb with real-time browser speech synthesis and a live telemetry dashboard (PR counts, CI uptime, sub-500ms protocol trace).
  • AWS Bedrock Reasoning: Ingests raw diagnostic JSON payloads from MCP tools and converts them into natural, conversational Alexa voice responses.

How we built it

  • Backend Protocol: Implemented in Python using FastAPI and Uvicorn, providing Streamable HTTP transport and endpoint handshakes conforming to MCP Spec 2025-11-25.
  • AWS Bedrock Integration: Integrated the Bedrock runtime using Boto3 to invoke foundation models (Claude/Llama) as the core reasoning engine.
  • Frontend & Ambient UI: A reactive dark-mode glassmorphic web dashboard with the Web Speech API and an animated audio-reactive ambient orb.
  • Repository Hygiene: Cleanly isolated with virtual environment guards, licensed under the MIT License for open-source community distribution.

Challenges we ran into

Handling raw streaming HTTP handshakes under the 2025-11-25 MCP specification required precise JSON schema structures. In early iterations, non-standard envelopes caused payload deserialization issues on chunked responses. We addressed this by enforcing strict Pydantic schemas across all tool endpoints.

Accomplishments that we're proud of

  • Sub-500ms end-to-end response loop: from intent query, through MCP tool execution and AWS Bedrock synthesis, back to voice output.
  • Live MCP protocol trace visible directly on the UI for transparent tool dispatching.
  • Complete 100% open-source, reproducible implementation requiring zero specialized physical hardware to simulate.

What we learned

  • How the Model Context Protocol (MCP) standardizes agent-to-tool handshakes.
  • Structuring prompt engineering for Amazon Bedrock to ensure foundation models produce voice-first, ambient-friendly answers instead of dense markdown.

What's next for OpsPulse

  • Deepening AWS AgentCore integration for autonomous multi-step rollbacks.
  • Adding physical Alexa+ smart display skills and Vega OS companion widgets.

Built With

Share this project:

Updates

Submission history