GenAI PMHarrison Chase
Founder and builder in the AI agents ecosystem, associated here with explaining deepagents’ architecture and its use of LangGraph. Relevant to PMs for understanding agent-loop/backend separation and deployment patterns.
Key Highlights
- Harrison Chase consistently frames agent systems as production architectures, not just prompts, with clear separation between reasoning loops and execution backends.
- His recent deepagents explanations are especially relevant to PMs evaluating local vs cloud deployment, sandboxing, MCP/a2a interfaces, and shared agent surfaces like Slack.
- He advocates model-agnostic harnesses and portable memory, giving PMs a practical framework for reducing vendor lock-in and preserving flexibility.
- His focus on traces, trajectories, evaluations, and observability provides a concrete playbook for improving agent quality over time.
- Managed Deep Agents and LangSmith Gateway reflect his broader push to simplify agent operations while adding governance, cost control, and infrastructure maturity.
Harrison Chase
Overview
Harrison Chase is a prominent builder in the AI agents ecosystem, best known for work associated with LangChain, LangGraph, LangSmith, and more recently deepagents and managed agent infrastructure. In the context captured here, he appears as a recurring explainer of how modern agent systems should be structured: separating the agent loop from execution backends, treating models as interchangeable components, and building around observability, evaluation, and deployment primitives rather than prompt logic alone.For AI Product Managers, Harrison Chase matters because his framing is highly operational. Instead of talking about agents as abstract autonomous systems, he emphasizes deployable architectures: model + harness + context, portable memory, backend separation for tools and code execution, and trace-driven improvement loops. These ideas are directly useful for PMs making decisions about agent product scope, infrastructure investment, reliability, vendor lock-in, and how to move from prototype agents to production systems.
Key Developments
- 2026-07-31: Unveiled LangSmith Gateway, highlighting cost controls, rate limiting, PII/data redaction, coding-agent integration, and access to open-source models such as kimi-k3.
- 2026-08-01: Shared internal agent evaluation benchmarks, standardizing on Harbor and measuring how well agents convert raw traces and data into Harbor tasks.
- 2026-08-05: Announced managed deepagents would enter public beta, focused on handling infrastructure so teams can prioritize agent logic; cited evaluations, memory, OAuth-based tool access, Slack/GitHub integrations, and sandbox support.
- 2026-08-08: Reflected on the path from early LangChain to managed agents, describing managed deepagents as a major launch and arguing managed agents can materially improve ease of running agent systems.
- 2026-08-09: Shared a longer explanation of Managed Deep Agents, describing it as a combination of a deep agent harness with managed LangSmith infrastructure; also reinforced the principle of separating the “brain” from the “hands.”
- 2026-08-11: Shared a tutorial on using Stagehand v4 with Managed Deep Agents and Browserbase to build a production-ready web-browsing agent.
- 2026-08-14: Shared a Max Agency episode with Unify CTO/co-founder Heggie Connor, highlighting major cost reductions before launch and a design pattern where subagents are effectively a function call; also emphasized low-friction model switching.
- 2026-08-15: Shared new LangSmith documentation on traces, threads, and trajectories, noting that observability data can support memory and learning when teams have a clear mental model.
- 2026-08-16: Recapped his “Owning Your Intelligence” talk, framing agents as model + harness + context and advocating portable memory, model-agnostic configurable harnesses, evaluations, observability, and a trace-driven data flywheel.
- 2026-08-17: Recapped deepagents’ architecture, explaining the separation of the agent loop from a backend that provides filesystem-like operations and optional sandboxed code execution; noted support for LangGraph, local or cloud deployments, MCP/a2a endpoints, shared web and Slack interfaces, and “fake” backends for non-coding agents.
Relevance to AI PMs
1. Agent architecture decisions: Harrison Chase’s framing helps PMs define clear boundaries between agent reasoning, tool execution, memory, and infrastructure. That is useful when deciding whether to build a thin orchestration layer, adopt a managed platform, or support multiple backends and deployment modes.2. Production-readiness and reliability: His emphasis on traces, trajectories, evaluations, and observability gives PMs a practical blueprint for turning agent experiments into measurable products. Teams can use these concepts to define success metrics, debugging workflows, and improvement loops.
3. Model and vendor strategy: His advocacy for model-agnostic harnesses, portable memory, and gateway-based controls is especially relevant for PMs trying to avoid vendor lock-in, manage cost, and switch models quickly without rewriting the whole product stack.
Related
- LangChain / LangGraph / LangSmith: Core parts of the ecosystem most closely associated with Harrison Chase; together they connect application logic, workflow orchestration, and observability.
- deepagents / Managed Deep Agents: The clearest recent expression of his architectural thinking, especially around separating agent loops from backends and making deployment easier.
- LangSmith Gateway / LangSmith Insights / traces / trajectories: Related to his repeated focus on observability, data flywheels, and infrastructure for monitoring and improving agents.
- Harbor / evaluations / agent reliability: Connect to his benchmark and evaluation thinking, which is highly relevant for PMs defining quality bars for agent products.
- MCP / a2a / Slack / GitHub / Browserbase / Stagehand v4: Examples of ecosystem interfaces and integrations that show how his work maps to real deployment surfaces and enterprise workflows.
- portable memory / model-neutrality / vendor lock-in / harnesses: Themes consistently tied to his product philosophy: keep the architecture configurable, observable, and portable across models and environments.
Newsletter Mentions (78)
“Harrison Chase recapped deepagents’ architecture, which separates the agent loop from a backend providing filesystem-like operations and, optionally, sandboxed code execution. Built on LangGraph, it supports local or cloud setups, standard endpoints such as MCP and a2a, shared web and Slack interfaces, and “fake” backends for non-coding agents.”
#1 𝕏 Harrison Chase recapped deepagents’ architecture, which separates the agent loop from a backend providing filesystem-like operations and, optionally, sandboxed code execution. Built on LangGraph, it supports local or cloud setups, standard endpoints such as MCP and a2a, shared web and Slack interfaces, and “fake” backends for non-coding agents.
“Harrison Chase recapped his “Owning Your Intelligence” talk, framing agents as model + harness + context and advocating portable memory plus model-agnostic, configurable harnesses. He also covered evaluations, observability, and a data flywheel that uses agent traces to drive improvement.”
#1 𝕏 Harrison Chase recapped his “Owning Your Intelligence” talk, framing agents as model + harness + context and advocating portable memory plus model-agnostic, configurable harnesses. He also covered evaluations, observability, and a data flywheel that uses agent traces to drive improvement.
“Harrison Chase shared new LangSmith documentation covering traces, threads, and trajectories—a new concept—and said observability data can support memory and learning, making a clear mental model especially helpful.”
#13 𝕏 Harrison Chase shared new LangSmith documentation covering traces, threads, and trajectories—a new concept—and said observability data can support memory and learning, making a clear mental model especially helpful.
“Harrison Chase shared a Max Agency episode with @unifygtm CTO and co-founder @HeggieConnor, covering how Unify cut costs by 90–95% two weeks before launch and why its subagents are just a function call.”
#13 𝕏 Harrison Chase shared a Max Agency episode with @unifygtm CTO and co-founder @HeggieConnor, covering how Unify cut costs by 90–95% two weeks before launch and why its subagents are just a function call. He said switching to @grok now takes one line of code.
“Harrison Chase shared a tutorial showing how to use Stagehand v4—an SDK that lets agents browse the web from Browserbase—with Managed Deep Agents to create a production-ready web browsing agent.”
Harrison Chase shared a tutorial showing how to use Stagehand v4—an SDK that lets agents browse the web from Browserbase—with Managed Deep Agents to create a production-ready web browsing agent.
“#3 𝕏 Harrison Chase shared a 20-minute explanation of Managed Deep Agents, which he said had launched the previous day and combines a deep agents harness with managed LangSmith infrastructure.”
GenAI PM Daily August 09, 2026 GenAI PM Daily 🎧 Listen to this brief 3 min listen Today's top 10 insights for PM Builders. Claude Code sessions can now message each other #3 𝕏 Harrison Chase shared a 20-minute explanation of Managed Deep Agents, which he said had launched the previous day and combines a deep agents harness with managed LangSmith infrastructure. He also planned to film more tutorials over the weekend. #4 𝕏 Harrison Chase shared Anthropic’s “Managed agents” engineering page, writing: “Separate brain from hands.”
“He said managed agents could significantly improve how easy it is to run agents.”
#4 𝕏 Harrison Chase shared his perspective on the journey from early LangChain to managed agents, describing managed deepagents as one of the launches he had been more excited about in a while. He said managed agents could significantly improve how easy it is to run agents.
“Harrison Chase announced that managed deepagents would move to public beta that week, focusing on infrastructure so users can prioritize agent logic.”
#5 𝕏 Harrison Chase announced that managed deepagents would move to public beta that week, focusing on infrastructure so users can prioritize agent logic. Features include Harbor-based evaluations, agent- and user-level memory, OAuth for tool access, Slack and GitHub integrations, and sandbox integration.
“Harrison Chase outlines his internal agent evaluation benchmarks, standardizing on Harbor and measuring each agent’s skill in converting raw traces and data into Harbor tasks.”
#12 𝕏 Harrison Chase outlines his internal agent evaluation benchmarks, standardizing on Harbor and measuring each agent’s skill in converting raw traces and data into Harbor tasks.
“Harrison Chase unveiled LangSmith Gateway, offering cost controls (including for end users), rate limiting, data/PII redaction, coding-agent integration, and access to OSS models like kimi-k3.”
#12 𝕏 Harrison Chase unveiled LangSmith Gateway, offering cost controls (including for end users), rate limiting, data/PII redaction, coding-agent integration, and access to OSS models like kimi-k3. #13 𝕏 Peter Yang got great feedback on his YouTube tutorial demonstrating how to use Claude to design and build a full-stack app end-to-end.
Related
An AI coding assistant environment used for running evaluation skills and agentic workflows. In this issue it is mentioned as a runtime for ai-evals-course material and as an agent in an OpenRouter-like system.
An AI company best known for Claude. It is referenced implicitly through Claude’s memory and Cowork features.
An AI coding tool referenced as providing data used to evaluate Grok 4.6. It is also named later as a target environment for running AI eval skills.
An AI coding agent or environment mentioned as a place to run AI eval skills. It is also listed as one of the agents that can be compared in a shared environment.
An AI infrastructure company and community that recapped a founder dinner in San Francisco. The discussion focused on vertical agents, moats, and go-to-market implications.
A standardized agent test suite referenced for model evaluation. The newsletter cites success rates on OpenClaw as part of the Nemotron benchmark result.
An interoperability protocol for connecting AI systems and tools. Here it is described through a public roadmap covering long-running workloads, local-server HTTP, discovery, identities, permissions, and generated SDKs.
A framework for building LLM applications and agents. In this newsletter it appears in the story about the founders’ attempt to automate dropshipping.
A handoff-oriented Claude workflow tool used to continue sessions and power inbox automation.
A workplace messaging platform used here as an operational surface for AI agents. PMs may care because agent integrations increasingly extend into team communication workflows.
A developer/evaluation tool cited in benchmark testing of automated eval systems. The newsletter uses it as part of a comparison against harder-to-detect product-judgment failures.
Autonomous or semi-autonomous AI systems that use tools, manage context, and complete tasks on behalf of users. The newsletter discusses common blockers such as tool quality, context overload, and system verification.
A software development platform used here as the source and sync target for repositories. It is central to AI coding workflows, plugin distribution, and agent automation.
An agent framework/architecture that separates the agent loop from backend operations like filesystem access and optional sandboxed code execution. Useful for building agents with flexible local/cloud deployment and multiple interfaces.
A 2.8T-parameter open-weight model described as frontier-level by the speaker in the newsletter. It is notable for strong quality and deployment on Nebius Token Factory.
A protocol or capability layer mentioned as part of an open, composable extension philosophy for AI tooling. It is grouped with MCP and Plugins.
An SDK for building Claude-based agents and workflows. It is cited as one of the newer harness-style tools replacing older frameworks.
Vercel’s SDK for integrating AI features into apps. The newsletter highlights token savings from a single line of code in DeepSeek-powered workflows.
Amazon’s cloud platform, referenced in a story about a training pipeline running on a 4-GPU instance. The anecdote highlights GPU utilization monitoring and infrastructure waste.
A framework for internal agents that emphasizes instructions, skills, channels, and connectors. It is presented as a default choice for building internal agent systems.
A configuration file used to steer agent behavior in repositories and workflows. Here it is used to configure CLI agents for a reusable research setup.
A standard endpoint/protocol for agent-to-agent or agent interoperability, mentioned here alongside MCP as a supported interface. It matters to PMs as part of agent connectivity and integration strategy.
A tool that provides coding agents with real-time API documentation so they can produce more accurate code. It targets agent-assisted development workflows.
A company focused on AI development workflows and agent harnesses. It is mentioned for its Missions framework and multi-step orchestration.
Community middleware example for customizing agent behavior and steering tasks in agent frameworks.
LangChain’s deployment offering for launching agents securely and at scale. It is important for PMs evaluating production readiness, observability, and managed infrastructure for agents.
A modular layer that adds tools, guardrails, and custom instructions to AI agents. It is described as a composable harness for production agent systems.
Stay updated on Harrison Chase
Get curated AI PM insights delivered daily — covering this and 1,000+ other sources.
Subscribe Free