Mezmo

Mezmo Mezmo engineers the context and orchestration foundation teams Spark insight with Mezmo.

Mezmo is the intelligence layer for production AI, pairing a production-grade control plane (AURA) with an optimized data plane (Mezmo) to make autonomous operations fast, efficient, and safe.

08/14/2026

Day 14 of 31 days of ✨AURA✨: One issue, two providers, zero custom code.

Henry Andrews investigates a live product catalog failure in Dash0's built-in OpenTelemetry demo, using AURA with two different providers stitched together: Dash0's hosted MCP server for telemetry and investigation tools, Baseten for model inference via a DeepSeek model.

Given only the symptom (product catalog is failing, front-end service is affected), AURA investigates through Dash0's live telemetry and returns a full report: impact, evidence, likely cause, what's still uncertain, and next steps.

No custom adapter required. Just AURA coordinating across whatever inference provider and MCP server you already use.

🎬 Watch day 14: https://www.mezmo.com/videos/no-custom-adapter-ai-sre-agent-aura-debugs-in-dash0

08/13/2026

Day 13 of 31 days of ✨AURA✨: From root cause to merged fix, without leaving the chat.

Continuing the homelab investigation: AURA now has GitHub access, letting it dig into the actual repos behind the broken smart lights instead of relying on logs alone.

Same root cause as before (a crash-looping container caused by a file ownership mismatch), but this time AURA recommends a concrete Docker Compose fix, drafts an issue with the root cause and proposed solution, and, after a human-in-the-loop approval, hands it off to Claude to open the PR.

Merge it, and AURA confirms the fix on its own: service healthy, root cause resolved. No manual log-diving, no host login required.

🎬 Watch day 13: https://www.mezmo.com/videos/from-log-line-to-merged-fix-ai-sre-agent-with-github-mcp

08/12/2026

Day 12 of 31 days of ✨AURA✨: A failed deployment. A short handoff. A fix, verified.

Henry Andrews sets up AURA as an Argo CD deployment guardian. When a sync fails, Argo sends AURA a short failure report via the agent-to-agent protocol, no details on cause or fix.

AURA investigates, finds a version mismatch between the deployment contract and what Argo expected, updates the one config map it's scoped to touch, and triggers a retry. Sync passes, application returns healthy.

Every model turn and tool call is traced in Phoenix, so you're not relying on AURA's final answer. You can see exactly what it did.

🎬 Watch day 12: https://www.mezmo.com/videos/argo-cd-deployment-failed-ai-sre-agent-aura-fixes-it

08/11/2026

Day 11 of 31 days of ✨AURA✨: A Lambda function that stopped running. AURA finds out why.

Greg Janco connects AURA to AWS using their MCP proxy and CLI, then asks it to debug a scheduled Lambda function that hasn't been running. AURA queries CloudWatch, inspects the function, and finds it's timing out at three seconds.

It confirms the finding through the logs, rules out a broader AWS Lambda outage, and recommends next steps to fix it.

No manual log-diving. Just a plain-language ask and a confirmed root cause.

🎬 Watch day 11: https://www.mezmo.com/videos/ai-sre-agent-debugs-a-lambda-timeout-via-aws-mcp-server

08/10/2026

Day 10 of 31 days of ✨AURA✨: No person, no alert, no approval step.

Henry Andrews has Kubernetes itself trigger AURA on a schedule, no human request or alert required. Every two minutes, a cron job spins up a fresh AURA instance, scoped to one namespace with exactly one allowed mutation: scaling the checkout deployment. When checkout scales to zero, AURA inspects the deployment, scales it back to 2 replicas, verifies the pods are healthy, and exits.

Guardrails are enforced at two levels: the system prompt limits AURA to a single bounded action, and the Kubernetes service account can only access the checkout deployment. Fully autonomous, fully scoped.

🎬 Watch day 10: https://www.mezmo.com/videos/scheduled-autonomous-ai-sre-agent-as-a-kubernetes-guardian

08/09/2026

Day 9 of 31 days of ✨AURA✨: Your runbooks, wherever you already keep them.

This demo connects AURA to Notion as a runbook database. Without it, a 3 am "storefront is crashing" page gets a generic, one-size-fits-all answer. Add a scoped runbook worker pointed at Notion, and AURA returns your actual documented steps, escalation contacts, and even when the runbook was last validated, clearly separating what came from your docs versus general knowledge.

One token, one config, no code. Your runbooks stay exactly where your team already stores them.

🎬 Watch day 9: https://www.mezmo.com/videos/ai-sre-agent-reads-your-notion-runbooks-in-an-incident

08/08/2026

Day 8 of 31 days of ✨AURA✨: The homelab strikes back.

This time, AURA is troubleshooting smart lights that Home Assistant can no longer control. It traces the issue to a crash-looping bridge, then a SQLite database stuck in read-only mode—all with zero prior knowledge of the environment.

When file permissions turn out fine, AURA's next move nails it: check what user the container runs as. Turns out it's running as a different user than the volume owner. Root cause, found.

No remediation yet, just the investigation.

🎬 Watch day 8: https://www.mezmo.com/videos/homelab-ai-sre-agent-debug-docker-container-permissions

08/07/2026

Day 7 of 31 days of ✨AURA✨: From "checkout is broken" to root cause, no hand-holding required.

Henry Andrews hands AURA a plain-language bug report: checkout is failing, browse and cart work fine, with no hints about which service or error to look for.

AURA queries live Mezmo telemetry, correlates checkout 500s with payment service warnings, and traces the issue to invalid payment token rejection, all while flagging exactly how confident it is in that correlation and why.

It doesn't just guess. It recommends concrete next steps and separates evidence from inference the whole way through.

🎬 Watch day 7: https://www.mezmo.com/videos/turn-a-slack-report-into-a-root-cause-with-an-ai-sre-agent

08/06/2026

Day 6 of 31 days of ✨AURA✨: So...why is this free?

Chip Johnson (and his dog, Willow) answers the question we keep getting: why give AURA away? The short version—Mezmo already built a world-class observability and OTEL platform, and it turns out that same data pipeline makes AURA faster, cheaper, and more accurate as an AI agent.

Giving AURA away means more people running AI agents in production, which grows the market for the enterprise-grade control plane we're building next.

AURA stays Apache 2.0, fully capable, and free—that part isn't changing.

🎬 Watch day 6: http://www.mezmo.com/videos/free-open-source-ai-agent-for-sre-and-more-the-aura-pledge

Address

2059 Camden Avenue #297
San Jose, CA
95124

Website

Alerts

Be the first to know and let us send you an email when Mezmo posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Mezmo:

Shortcuts

Share

Category