Summaries of every session on the MLOps Community YouTube channel.https://mlopstalks.comen-gbhttps://mlopstalks.com/talks/agentic-dx-bringing-idp-into-your-idehttps://mlopstalks.com/talks/agentic-dx-bringing-idp-into-your-ideAdnan Vahora describes how Motorola Solutions tried to increase adoption of an Internal Developer Platform used by more than 4,000 monthly active users. The platform had only 22% adoption, so the team moved its agent experience into the IDE through a Bring Your Own Agent approach. Early designs exposed problems with duplicated engineering work, flat agent meshes, routing, hallucinations, authentication,...Wed, 19 Aug 2026 00:00:00 GMTagentsdeveloper-experiencemcpplatform-teamshttps://mlopstalks.com/talks/before-the-agent-calls-source-level-findings-from-100-mcp-servershttps://mlopstalks.com/talks/before-the-agent-calls-source-level-findings-from-100-mcp-serversAkash Sathish examines MCP security before an agent makes its first tool call. He describes how malicious instructions can enter through prompts, connected MCP servers, or retrieved documents, then shows how an LLM can steer an MCP server toward dangerous behavior. His audit of 100 widely installed TypeScript MCP servers found 445 findings across 70 servers. The examples include command...Wed, 19 Aug 2026 00:00:00 GMTmcpsecurityhttps://mlopstalks.com/talks/mcp-2-0-what-changes-what-breaks-and-what-you-need-to-do-about-ithttps://mlopstalks.com/talks/mcp-2-0-what-changes-what-breaks-and-what-you-need-to-do-about-itMateo Torres explains MCP 728 as the largest rewrite of the protocol since its launch. The central change is a move from stateful sessions to stateless requests. Requests carry the context a server needs, so any replica can handle them and a failed replica does not take the session with it. This simplifies server scaling, although it makes client implementations...Wed, 19 Aug 2026 00:00:00 GMTdeploymentlatencymcphttps://mlopstalks.com/talks/mcp-release-overview-stateless-and-the-big-changes-in-the-new-spechttps://mlopstalks.com/talks/mcp-release-overview-stateless-and-the-big-changes-in-the-new-specAlex Hancock explains the largest MCP protocol change since its launch. MCP started with local hosts and data sources, but more usage now happens between remote clients and servers. The new design removes sessions and initialization, so each request carries the information a server instance needs. This makes horizontal scaling easier because requests can move between instances without session pinning...Wed, 19 Aug 2026 00:00:00 GMTdeploymentmcpsecurityhttps://mlopstalks.com/talks/mcps-for-observability-stackshttps://mlopstalks.com/talks/mcps-for-observability-stacksDiana Todea shows how Model Context Protocol can connect an AI assistant to an observability stack. Her examples combine a VictoriaMetrics cloud deployment with a separate anomaly detection MCP server. The assistant retrieves metrics and logs, reviews metric usage, searches documentation, selects anomaly detection models, and drafts alert configurations. Todea first explains the value of bringing metrics, logs, traces, and...Wed, 19 Aug 2026 00:00:00 GMThuman-in-the-loopmcpmonitoringobservabilityhttps://mlopstalks.com/talks/responsible-autonomy-building-governance-frameworks-for-ai-that-act-in-the-realhttps://mlopstalks.com/talks/responsible-autonomy-building-governance-frameworks-for-ai-that-act-in-the-realSaurabh Mishra explains how AI has moved from predictive models and chatbots toward agents that can reason, use tools, and perform actions in business systems. MCP provides a standard way for agents to communicate with tools and services, but this creates risks when agents can change infrastructure, access CRM or payroll systems, or trigger other workflows. Mishra presents governance as...Wed, 19 Aug 2026 00:00:00 GMTgovernancemcpobservabilitysecurityhttps://mlopstalks.com/talks/stateless-yet-durable-mcp-tasks-v2https://mlopstalks.com/talks/stateless-yet-durable-mcp-tasks-v2Cornelia Davis explains MCP Tasks v2 through a purchase order workflow that records received goods, updates inventory, closes the purchase order, and processes an invoice. The invoice may take seconds, days, or weeks because it can wait for approvals, policy checks, backend checks, and user input. MCP tools can run synchronously or asynchronously, with asynchronous calls returning a task handle....Wed, 19 Aug 2026 00:00:00 GMTmcpworkflowshttps://mlopstalks.com/talks/the-mcp-tasks-extensionhttps://mlopstalks.com/talks/the-mcp-tasks-extensionVikram Vaswani explains why synchronous Model Context Protocol requests are a poor fit for long-running work. Batch migrations, report generation, model training, and multi-step agent workflows may need progress checks, cancellation, or input from a user while they run. The earlier Tasks design supported these cases, but it required capability checks in several places, used a blocking task/result call for...Wed, 19 Aug 2026 00:00:00 GMTmcpworkflowshttps://mlopstalks.com/talks/what-we-learned-from-dozens-of-enterprise-mcp-deploymentshttps://mlopstalks.com/talks/what-we-learned-from-dozens-of-enterprise-mcp-deploymentsJiquan Ngiam shares lessons from working with enterprise teams that deploy MCP and agents in production. He argues that the main problem has moved from agent intelligence to getting governed data into agents safely. Everyday work tools such as Gmail, Slack, Figma, HubSpot, and Airtable are driving much of the usage, including among non-engineers. Enterprise connections are harder than individual...Wed, 19 Aug 2026 00:00:00 GMTgovernancemcpsecuritytool-usehttps://mlopstalks.com/talks/events-notifications-in-mcphttps://mlopstalks.com/talks/events-notifications-in-mcpAman Singh presents a design sketch for Events, a proposed Model Context Protocol extension. Current MCP notifications require an open connection and usually carry only a resource URI, so the client must fetch the changed resource separately. Events would let an agent subscribe with a name and parameters, then receive a structured event payload. The proposal includes polling with cursors,...Mon, 17 Aug 2026 00:00:00 GMTmcptool-useworkflowshttps://mlopstalks.com/talks/json-schema-2020-12-and-the-contract-for-contexthttps://mlopstalks.com/talks/json-schema-2020-12-and-the-contract-for-contextOla Hungerford explains why MCP moved from a small JSON Schema subset to JSON Schema 2020-12 for tool schemas. The earlier limits caused SDKs and servers to rewrite schemas, wrap output values in objects, and add validation outside the protocol. The updated specification keeps the object requirement for input schemas but allows the rest of the dialect, including conditional keywords...Mon, 17 Aug 2026 00:00:00 GMTmcpstructured-outputstestingtool-usehttps://mlopstalks.com/talks/loop-engineeringhttps://mlopstalks.com/talks/loop-engineeringThis reading group discusses loop engineering rather than walking through a named paper or book. Sparsh Jain frames it as the next stage after prompt, context, and harness engineering. A loop discovers work, plans it, delegates tasks, verifies changes, stores state, and repeats until a goal or stopping condition is reached. Arthur Coleman describes building a production platform with Claude...Mon, 17 Aug 2026 00:00:00 GMTagentscoding-agentsevalsobservabilityworkflowshttps://mlopstalks.com/talks/mcp-goes-statelesshttps://mlopstalks.com/talks/mcp-goes-statelessJohn Dellenbaugh and Pankaj Kumar explain why the 2026-07-28 MCP specification changes horizontal scaling. Their shopping-cart demo first shows the old design, where an MCP server keeps session data in local memory. When a load balancer sends later requests to another instance, the server returns "session not found". Sticky routing fixes request placement, while a session store preserves data after...Mon, 17 Aug 2026 00:00:00 GMTdeploymentmcphttps://mlopstalks.com/talks/policy-enforcement-and-tamper-evident-audit-chainshttps://mlopstalks.com/talks/policy-enforcement-and-tamper-evident-audit-chainsImran Siddique argues that MCP adoption has moved beyond basic agents and chatbots, while policy enforcement has become a practical requirement for production systems. Policy alone does not prove that the policy stayed unchanged or that an agent followed it. An administrator, agent, or compromised system could alter the policy, decision, or audit log. His confidential MCP design places the...Mon, 17 Aug 2026 00:00:00 GMTgovernancemcpsecurityhttps://mlopstalks.com/talks/coding-agents-are-secretly-general-agentshttps://mlopstalks.com/talks/coding-agents-are-secretly-general-agentsJay Hack traces coding agents from early tools such as Kite and GitHub Copilot to Cursor and autonomous ticket-to-pull-request systems. He explains why code became a strong training ground: generated code can compile, pass tests, or be measured for performance, giving agents feedback they can act on. From there, he argues that coding agents are general agents because code is...Thu, 23 Jul 2026 00:00:00 GMTagentscoding-agentsevalsreinforcement-learningtool-usehttps://mlopstalks.com/talks/sandboxing-agent-harnesses-and-agent-teamworkhttps://mlopstalks.com/talks/sandboxing-agent-harnesses-and-agent-teamworkShahram Anver explains how Cleric changed its AI SRE architecture as models improved. The company moved from a complicated ant-colony design with many specialized workers to a simpler query agent, then relaxed rigid tool restrictions as models became better at composing tools. That change required strong sandboxing because bash and Python give an agent much more freedom. Anver argues that...Mon, 20 Jul 2026 00:00:00 GMTagentsguardrailsmemorymulti-agentreliabilityhttps://mlopstalks.com/talks/omnigent-composition-control-and-collaboration-for-ai-agentshttps://mlopstalks.com/talks/omnigent-composition-control-and-collaboration-for-ai-agentsDenny Lee describes Omnigent as an open-source layer above models and agent harnesses. He wants developers to keep their context, skills, memory, and configuration while switching between systems such as Claude Code, Codex, Pi, and other models. His matcha research in Taiwan is the practical example: two agents debate regions, soil, elevation, language sources, farmers, and processing requirements, which produces...Wed, 15 Jul 2026 00:00:00 GMTagentsguardrailsmemorymulti-agenthttps://mlopstalks.com/talks/logs-are-all-you-need-rethinking-observability-with-ai-agentshttps://mlopstalks.com/talks/logs-are-all-you-need-rethinking-observability-with-ai-agentsSherwood Callaway describes Sazabi as an AI-native observability platform for teams that ship quickly. Its interface is chat, including a Slackbot, because engineers can ask an agent what went wrong instead of inspecting telemetry directly. Sazabi focuses on logs and reconstructs metrics and traces from them, which removes the need to configure Prometheus, understand metric types, or propagate trace context....Mon, 22 Jun 2026 00:00:00 GMTagentsevalsobservabilitysecurityhttps://mlopstalks.com/talks/autonomous-agents-at-work-from-openclaw-hype-to-enterprise-realityhttps://mlopstalks.com/talks/autonomous-agents-at-work-from-openclaw-hype-to-enterprise-realityPramod Krishnan explains why agents that take actions need stronger controls than systems that only answer questions. He classifies work as reversible, sensitive, or consequential, then maps each category to different approval and testing requirements. His production framework covers agent identity and credentials, prompt-injection defenses, allow-listed tools, output limits, rollback paths, audit logs, evaluations, and cost controls. He describes auditability...Fri, 19 Jun 2026 00:00:00 GMTagentscostgovernanceguardrailshttps://mlopstalks.com/talks/voice-agent-use-caseshttps://mlopstalks.com/talks/voice-agent-use-casesAnurag Beniwal describes the engineering choices behind production voice agents. He prefers architectures between simple cascaded pipelines and fully speech-to-speech systems because enterprises need control over models, tools, latency, and failures. A smaller model can keep a caller engaged while retrieval, tool calls, or a more expensive model work in the background. He explains why voice adds problems that chat...Fri, 19 Jun 2026 00:00:00 GMTagentslatencytool-usevoicehttps://mlopstalks.com/talks/agents-the-40m-bet-on-multiplayer-aihttps://mlopstalks.com/talks/agents-the-40m-bet-on-multiplayer-aiStanislas Polu describes the shift from single-player AI, where one person works privately with an agent, toward multiplayer AI for work that spans teams and lasts days or weeks. Dust's pod model groups humans, agents, sessions, tasks, and shared files around one project. He explains how this can automate a team weekly by creating sessions for each slide, gathering company...Mon, 15 Jun 2026 00:00:00 GMTagentscostgovernancemulti-agentworkflowshttps://mlopstalks.com/talks/context-engineering-for-coding-agentshttps://mlopstalks.com/talks/context-engineering-for-coding-agentsFausto presents coding agents as systems whose behaviour depends heavily on context injection. Engineers cannot directly control the model's internal parameters, but they can control what enters its context window and when. He discusses Claude Code files, rules, hooks, sub-agents, skills, plugins, observers, and deferred tools. His practical advice is to keep context small, use project-scoped information, and start a...Mon, 15 Jun 2026 00:00:00 GMTagentscoding-agentscontext-engineeringmemorytool-usehttps://mlopstalks.com/talks/ai-is-fast-ai-projects-are-slow-lets-fix-thathttps://mlopstalks.com/talks/ai-is-fast-ai-projects-are-slow-lets-fix-thatJoe Maionchi and Rod Christensen describe RocketRide as an open-source framework for building AI pipelines from typed nodes connected by data lanes. Their argument is that coding agents can write application code quickly, but they often produce inconsistent infrastructure code, lose context, miss reusable components, and mishandle asynchronous execution. RocketRide defines the plumbing around models, tools, agents, and data so...Tue, 09 Jun 2026 00:00:00 GMTinferenceobservabilityopen-sourceworkflowshttps://mlopstalks.com/talks/architecting-modern-ai-systemshttps://mlopstalks.com/talks/architecting-modern-ai-systemsThe panel discusses how AI platforms are changing as teams adopt agents, open-source models, and more distributed systems. Frederick describes a mental-health hackathon run with Bell Canada and Kids Help Phone, where the evaluation pipeline ran on Kubernetes in the BuzzHPC environment and handled more than 1,000 submissions. The speakers compare managed APIs with self-hosted models, covering cost, data residency,...Thu, 28 May 2026 00:00:00 GMTagentsevalsgovernanceobservabilityhttps://mlopstalks.com/talks/architecting-modern-ai-systems-platforms-agents-and-integrationhttps://mlopstalks.com/talks/architecting-modern-ai-systems-platforms-agents-and-integrationThe panel discusses how AI platforms are changing as teams adopt agents, open models, and more distributed systems. The hackathon example grounds the discussion: more than 100 teams built mental-health conversational systems, while the organisers ran evaluation infrastructure on Kubernetes with hosted models, GPUs, CPUs, and a leaderboard. The speakers describe why teams may move from API-based models to locally...Thu, 28 May 2026 00:00:00 GMTagentsevalsopen-modelsplatform-teamshttps://mlopstalks.com/talks/whats-special-about-metas-multi-agent-systemshttps://mlopstalks.com/talks/whats-special-about-metas-multi-agent-systemsAditya Gautam describes an application-layer multi-agent system for two short-form video problems: modality misalignment and original-content attribution. A video can contain a brief policy-breaking segment that conflicts with its surrounding content, or a creator can re-upload someone else's work with small edits. Meta's design separates the work into a Perceiver that extracts video signals, a Retriever that searches indexed content...Mon, 25 May 2026 00:00:00 GMTagentscostmulti-agentmultimodalsearchhttps://mlopstalks.com/talks/building-mcp-before-mcp-existed-inside-despegars-sofia-agenthttps://mlopstalks.com/talks/building-mcp-before-mcp-existed-inside-despegars-sofia-agentNicolas Alejandro Bogliolo explains how Despegar built Sofia, its travel concierge, before LangChain and the Model Context Protocol were widely available. The system began with one assistant calling tools directly. As the toolset grew, Despegar added Chappie as an orchestration layer and moved specialist agents with their tools outside it. Flights, hotels, activities, and other areas can be owned by...Wed, 20 May 2026 00:00:00 GMTagentsmcpmulti-agenttool-useworkflowshttps://mlopstalks.com/talks/the-latency-goldilocks-zone-explainedhttps://mlopstalks.com/talks/the-latency-goldilocks-zone-explainedRafael Borger and Daniel Wolbert explain how iFood is building ILO, a conversational ordering experience for users with very different tastes, budgets, and habits. ILO can interpret requests such as a dinner for two where one person is vegetarian, then combine many signals to select suitable options. The system also tries to help undecided users through recommendations and a swipe-based...Wed, 20 May 2026 00:00:00 GMTagentsrecommender-systemssearchvoicehttps://mlopstalks.com/talks/building-ai-agents-that-survive-productionhttps://mlopstalks.com/talks/building-ai-agents-that-survive-productionHaytham Abuelfutuh argues that production agents will fail, so teams should make those failures cheap to handle. He presents three design principles for an agent platform: Dynamic, Durable, and Defended. Dynamic means developers can write ordinary Python with loops, branching, and error handling instead of fitting an agent into a constrained DSL. Durable means the platform records actions, caches non-deterministic...Thu, 14 May 2026 00:00:00 GMTagentscachingguardrailsreliabilityhttps://mlopstalks.com/talks/stop-ai-agents-from-sql-injecting-your-databasehttps://mlopstalks.com/talks/stop-ai-agents-from-sql-injecting-your-databaseAveri Kitsch explains what Google has learned from more than 20 million monthly tool calls to Google Cloud databases. She describes three common database-tool patterns, then separates flexible build-time assistants from constrained runtime tools that serve untrusted users. The security problem is an agent with private data, untrusted content, and a way to send results back. Kitsch demonstrates how a...Mon, 11 May 2026 00:00:00 GMTmcpsecuritytool-usehttps://mlopstalks.com/talks/why-ai-agents-shouldnt-replace-your-fraud-modelshttps://mlopstalks.com/talks/why-ai-agents-shouldnt-replace-your-fraud-modelsVarant Zanoyan argues that agents should improve the models behind fraud detection, underwriting, search ranking, and other high-stakes systems instead of replacing those models at decision time. These systems need auditability, low latency, high throughput, and stable performance. His proposed pattern is agentic experimentation: an agent creates features, trains and evaluates a model, and deploys a complete pipeline to dev....Mon, 11 May 2026 00:00:00 GMTagentsdata-pipelinesdeploymentfeature-engineeringhttps://mlopstalks.com/talks/the-creator-of-superpowers-why-real-agentic-engineering-beats-vibe-codinghttps://mlopstalks.com/talks/the-creator-of-superpowers-why-real-agentic-engineering-beats-vibe-codingJesse Vincent explains how Superpowers turns Claude Code into a more disciplined software development workflow. The process starts with brainstorming, where the agent helps the human work out what they actually want. A planning skill then creates small tasks with explicit files, tests, success conditions, and implementation guidance. Separate sub-agents implement each task, check spec compliance, and review code quality....Tue, 05 May 2026 00:00:00 GMTagentscoding-agentsmulti-agenttestingworkflowshttps://mlopstalks.com/talks/getting-humans-out-of-the-way-how-to-work-with-teams-of-agentshttps://mlopstalks.com/talks/getting-humans-out-of-the-way-how-to-work-with-teams-of-agentsRob Ennals describes a way to work with coding agents without becoming the reviewer for every line and interaction. His feature walk-through process has an agent produce cropped screenshots and explanations, then asks another agent to verify them. The result gives the human a fast review path and can also support visual regression checks. Ennals applies the same principle to...Fri, 01 May 2026 00:00:00 GMTagentscoding-agentsmulti-agentobservabilitytestinghttps://mlopstalks.com/talks/openxdata-conferencehttps://mlopstalks.com/talks/openxdata-conferenceOpenXData Conference 2026 covers how open data systems are changing as AI agents become major consumers of data. The keynote argues that lakehouses must evolve from systems of record into context-serving layers that provide fresh, joinable, governed data at low latency. Other sessions address guardrails for generated SQL and Spark jobs, Apache Hudi optimizations at JD.com and Uber, online feature...Wed, 29 Apr 2026 00:00:00 GMTcontext-engineeringdata-engineeringfeature-engineeringfeature-storesguardrailshttps://mlopstalks.com/talks/its-2026-and-were-still-talking-evalshttps://mlopstalks.com/talks/its-2026-and-were-still-talking-evalsMaggie Konstanty argues that evals should begin when a product idea begins and continue throughout development and production. Pre-ship tests can simulate personas and scenarios, but real users introduce requests and failure modes that teams did not imagine. She prefers evaluating specific failures, such as recommending pepperoni to a vegetarian user, over relying on an unexplained accuracy percentage. For food...Mon, 27 Apr 2026 00:00:00 GMTevalsmonitoringobservabilitytestinghttps://mlopstalks.com/talks/why-agents-are-driving-software-development-to-the-cloudhttps://mlopstalks.com/talks/why-agents-are-driving-software-development-to-the-cloudZach Lloyd argues that coding agents should move off individual laptops and into a shared cloud system. He compares agents to teammates or cloud functions, each with its own permissions, workspace, memory, and task. Warp's Oz platform records agent plans, conversations, verification steps, and outputs so teams can review work, reuse knowledge, steer agents, and keep data in company-controlled infrastructure....Mon, 27 Apr 2026 00:00:00 GMTagentsmulti-agentobservabilitytool-usehttps://mlopstalks.com/talks/a-new-kind-of-marketplacehttps://mlopstalks.com/talks/a-new-kind-of-marketplaceDonné Stevenson and Pedro Chaves discuss how agents could change buying and selling across real estate, automotive, and general classified marketplaces. Pedro describes a real-estate experience that treats house hunting as a lifestyle decision. An agent gathers preferences about neighborhoods, amenities, schools, safety, and commuting, then combines chat with clickable components and enriched recommendations. For dealers, Donné describes an early...Mon, 20 Apr 2026 00:00:00 GMTagentsecommerceguardrailsmulti-agenthttps://mlopstalks.com/talks/how-we-cut-llm-latency-70-with-tensorrt-in-productionhttps://mlopstalks.com/talks/how-we-cut-llm-latency-70-with-tensorrt-in-productionMaher Hanafi describes how he moved from having no AI experience to leading self-hosted LLM infrastructure for an HR technology company. He explains the hidden work behind AI features, including GPU selection, latency, throughput, cost, cold starts, model quality, privacy, and language support. His team used AWS FSx and container images with embedded models to reduce cold-start time. TensorRT LLM...Mon, 20 Apr 2026 00:00:00 GMTcostgpusinferenceplatform-teamshttps://mlopstalks.com/talks/the-modern-software-engineerhttps://mlopstalks.com/talks/the-modern-software-engineerMihail Eric describes software engineering in a world where coding agents can write, test, and explain code. He argues that agents require more upfront thought because engineers must define architecture, split work, delegate tasks, and create checks that show whether an agent has gone wrong. He is skeptical of social-media demonstrations with dozens of agents running at once. In his...Mon, 20 Apr 2026 00:00:00 GMTagentscoding-agentsplanningtestinghttps://mlopstalks.com/talks/fixing-gpu-starvation-in-large-scale-distributed-traininghttps://mlopstalks.com/talks/fixing-gpu-starvation-in-large-scale-distributed-trainingKashish Mittal describes GPU starvation as a data pipeline problem. Models can sit idle because training reads too much data, moves duplicated features between CPU and GPU hosts, or spends time converting data into a format the GPU can use. At Uber, models running on A100 chips reached only 15% to 20% utilization. Loading the data directly into RAM raised...Fri, 10 Apr 2026 00:00:00 GMTcachingdata-pipelinesgpusinferencehttps://mlopstalks.com/talks/ship-agents-a-virtual-conference-track-2https://mlopstalks.com/talks/ship-agents-a-virtual-conference-track-2Track 2 covers four production problems for agents. Sarmad Absil presents a manufacturing system that turns IoT anomalies into shared Microsoft Teams alerts, then uses specialized agents to retrieve maintenance history, inventory, manuals, and work-order data. The system keeps humans responsible for physical repairs. Kamal presents a security workflow that moves threat modeling and software supply-chain checks into the developer...Fri, 10 Apr 2026 00:00:00 GMTagentsinferenceobservabilitysecurityhttps://mlopstalks.com/talks/mcp-dev-summit-day-2https://mlopstalks.com/talks/mcp-dev-summit-day-2Day 2 of MCP Dev Summit focused on how MCP is moving beyond simple tool calls. Ido Salomon and Lead Ysef presented MCP apps, which let servers return interactive HTML-based interfaces that preserve a company's branding and let users act through the host. Ryan Cook described WorkOS's context engine, which resolves identity, permissions, integration instructions, and semantic definitions before an...Fri, 03 Apr 2026 00:00:00 GMTcontext-engineeringgovernancemcpsecurityhttps://mlopstalks.com/talks/the-coding-agent-multiverse-of-madnesshttps://mlopstalks.com/talks/the-coding-agent-multiverse-of-madnessAnkit Mathur argues that enterprises are losing control as developers adopt Claude Code, Cursor, Codex, and other coding agents for different tasks. IT teams face separate vendor contracts, scattered billing, limited visibility, and difficult security policies around sensitive data. Databricks built a coding agent gateway to put these tools behind one access point with shared reporting, cost limits, privacy controls,...Fri, 03 Apr 2026 00:00:00 GMTagentscoding-agentscostgovernanceobservabilityhttps://mlopstalks.com/talks/mcp-dev-summit-day-1https://mlopstalks.com/talks/mcp-dev-summit-day-1Day 1 presents MCP as an integration layer that is moving from rapid adoption into enterprise production. David Soria Parra describes the protocol's growth from local stdio servers to remote systems, authorization, structured outputs, tasks, extensions, and MCP Apps. His roadmap includes stateless transport, better long-running task support, enterprise authorization, triggers, streaming, and skills. Builders from Amazon, Uber, and Duolingo...Thu, 02 Apr 2026 00:00:00 GMTgovernancemcpobservabilitysecurityhttps://mlopstalks.com/talks/2026-the-year-of-agent-orchestrationhttps://mlopstalks.com/talks/2026-the-year-of-agent-orchestrationZach Lloyd argues that coding has shifted from writing code by hand to writing prompts and guiding agents. As engineers run several agents at once, laptops become a limiting factor, while team leaders lose visibility into the work, security exposure, and agent availability. Lloyd expects teams to move agents into the cloud so they can run longer tasks, work in...Tue, 31 Mar 2026 00:00:00 GMTagentscoding-agentshuman-in-the-loopmulti-agentworkflowshttps://mlopstalks.com/talks/decomposing-the-agent-orchestration-system-lessons-learnedhttps://mlopstalks.com/talks/decomposing-the-agent-orchestration-system-lessons-learnedNiels Bantilan explains what Union learned while building Flyte 2.0 and productionizing agentic systems. His concern is the infrastructure around an agent, especially failures caused by resource limits, preemptions, networking, lost state, and parallel work. He proposes six design principles: use a general-purpose language such as Python, add durability and observability hooks, make failures cheap, expose infrastructure as context, give...Tue, 31 Mar 2026 00:00:00 GMTagentscachingmulti-agentreliabilityhttps://mlopstalks.com/talks/how-to-make-a-coding-agent-a-general-purpose-agenthttps://mlopstalks.com/talks/how-to-make-a-coding-agent-a-general-purpose-agentHarrison Chase and Sam Partee divide a general purpose agent into two layers. The agent harness gives a model a coding-agent experience through file tools, planning, subagents, skills, context management, persistence, and human approval. The tool runtime connects that agent to services outside its local environment. It handles authentication, authorization, delegated permissions, enterprise identity systems, secrets, and integrations such as...Tue, 31 Mar 2026 00:00:00 GMTagentscoding-agentscontext-engineeringhuman-in-the-looptool-usehttps://mlopstalks.com/talks/stop-shipping-on-vibes-how-to-build-real-evals-for-coding-agentshttps://mlopstalks.com/talks/stop-shipping-on-vibes-how-to-build-real-evals-for-coding-agentsJessica Wang argues that teams often ship AI features because a PM tried a few prompts or an engineer declared the feature ready. She proposes using evals to make those decisions measurable. An eval needs test data, a defined task, a scoring method, and experiment runs that can be compared for regressions, quality changes, cost, and latency. Wang then walks...Tue, 31 Mar 2026 00:00:00 GMTagentscoding-agentsevalssearchhttps://mlopstalks.com/talks/everything-we-got-wrong-about-research-plan-implementhttps://mlopstalks.com/talks/everything-we-got-wrong-about-research-plan-implementDexter Horthy revisits the Research-Plan-Implement workflow after seeing experts get good results while their teams struggled. Research often became opinionated because agents were told what feature they were building. Planning also failed when a large prompt skipped its questions and design discussion, often because the workflow exceeded the model's instruction budget. Horthy now separates the process into questions, research, design,...Tue, 24 Mar 2026 00:00:00 GMTagentscoding-agentscontext-engineeringhuman-in-the-loopplanninghttps://mlopstalks.com/talks/lessons-from-25-trillion-tokens-scaling-ai-assisted-development-at-kilohttps://mlopstalks.com/talks/lessons-from-25-trillion-tokens-scaling-ai-assisted-development-at-kiloScott Breitenother describes how Kilo changed its engineering model after processing 25 trillion tokens and serving more than 1.5 million developers. The company has about 15 engineers, each owning features from design through deployment and user feedback. Developers increasingly act as orchestrators, while agents handle coding, documentation, debugging, and other execution work. Scott says the transition happens gradually, from autocomplete...Tue, 24 Mar 2026 00:00:00 GMTagentscoding-agentscontext-engineeringmulti-agentteam-adoption