GenAI PMHarrison Chase
Founder and CEO of LangChain, known for agent and orchestration commentary. Here he argues that agent improvement depends on harness design, tool boundaries, and measurement.
Key Highlights
- Harrison Chase argues that agent improvement is primarily harness improvement, especially at the tool boundary.
- He frames production agents as a combination of model, harness, and context rather than model quality alone.
- His recent commentary emphasizes portable memory, model-agnostic architecture, and lower vendor lock-in.
- He connects LangSmith observability data to debugging, evaluations, memory, and continuous improvement loops.
- He has highlighted deepagents as an architecture that separates orchestration logic from swappable backend infrastructure.
Harrison Chase
Overview
Harrison Chase is the founder and CEO of LangChain and a widely cited voice on how to build reliable AI agents in production. In recent commentary, he has consistently argued that agent quality is not just a model problem; it is a harness problem. His framing emphasizes the interaction between the model, the harness, and the context layer, with improvement coming from better orchestration, tool boundaries, memory, observability, and evaluation.For AI Product Managers, Harrison Chase matters because his work sits at the junction of product architecture and operational reliability. Across LangChain, LangGraph, Deep Agents, and LangSmith, he has pushed a practical worldview: durable workspaces, inspectable traces, portable memory, model-agnostic harnesses, and measurable recovery loops are what make agent systems shippable. That perspective is especially relevant for PMs deciding how to reduce vendor lock-in, compare models, instrument agent behavior, and improve task success over time.
Key Developments
- 2026-08-09: Harrison Chase shared a 20-minute explanation of Managed Deep Agents, describing it as a deep agents harness combined with managed LangSmith infrastructure. He also highlighted Anthropic’s managed agents work with the framing “Separate brain from hands.”
- 2026-08-11: He shared a tutorial showing how Stagehand v4 from Browserbase can be used with Managed Deep Agents to build a production-ready web-browsing agent.
- 2026-08-14: He shared a Max Agency episode with Unify CTO and co-founder Heggie Connor, discussing large cost reductions before launch and an architecture where subagents can be treated as a function call. He also emphasized how easily model providers can be swapped.
- 2026-08-15: He shared new LangSmith documentation covering traces, threads, and trajectories, and argued that observability data can also power memory and learning when teams have a clean mental model.
- 2026-08-16: He recapped his “Owning Your Intelligence” talk, defining agents as model + harness + context. He advocated portable memory, model-agnostic configurable harnesses, evaluations, observability, and a data flywheel built from agent traces.
- 2026-08-17: He recapped the deepagents architecture, separating the agent loop from a backend that provides filesystem-like operations and optional sandboxed code execution. He noted support for local or cloud setups, LangGraph foundations, standards like MCP and A2A, and shared web or Slack interfaces.
- 2026-08-30: He argued that model labs will build harnesses for their own models while limiting access to rival harnesses, making a lab-independent harness the main path for teams that want cross-model portability.
- 2026-09-03: He shared that LangSmith’s Messages View helps teams debug agents by replaying conversations and tool calls from the agent’s perspective, making trace data useful to builders and not just infrastructure teams.
- 2026-09-04: He said agent workspaces should be durable, inspectable, and swappable, pointing to MongoDB support for LangChain Deep Agents’ virtual file system via BackendProtocol with operations like read, write, glob, and grep.
- 2026-09-10: He endorsed the idea that improving agents means improving the harness, especially at the tool boundary through context selection, tool availability, failure recovery, and measurement. He described a loop being built around deepagents for orchestration logic and LangSmith for measurement.
Relevance to AI PMs
1. Use harness thinking to improve agent outcomes. Harrison Chase’s model suggests that when agents underperform, PMs should inspect context selection, available tools, fallback logic, and recovery loops before assuming the base model is the main issue. This gives teams a concrete roadmap for raising task completion and reliability.2. Design for portability and lower vendor lock-in. His emphasis on model-agnostic harnesses, portable memory, and swappable backends is directly useful for PMs choosing stack strategy. It supports procurement flexibility, easier model comparison, and safer transitions between labs or infrastructure vendors.
3. Make observability part of the product loop. His focus on traces, trajectories, messages views, and measurement shows PMs how to treat observability as a product improvement system—not just an engineering dashboard. This is useful for prioritizing evals, understanding failure modes, and feeding memory or learning systems.
Related
- LangChain: The company and ecosystem Harrison Chase leads; central to his thinking on model orchestration and production agent systems.
- LangGraph: The orchestration foundation mentioned in deepagents architecture, relevant for building stateful multi-step agents.
- LangSmith: The measurement and observability layer repeatedly tied to traces, messages, trajectories, and production debugging.
- deepagents / Managed Deep Agents: Closely associated with his recent commentary on separating the agent loop from the backend, adding managed infrastructure, and improving agents through harness design.
- agent-harnesses / model-agnostic-harnesses / open-harnesses: Concepts strongly connected to his thesis that agent performance depends on orchestration logic more than model choice alone.
- memory / portable-memory: Connected to his “Owning Your Intelligence” framing, where long-term value comes from memory systems that can move across models and products.
- observability / evaluations / agent-traces / trajectories: Core to his proposed improvement loop for production agents.
- MCP / A2A / BackendProtocol: Standards and interfaces that connect to his emphasis on swappable backends, interoperable tooling, and durable workspaces.
- Browserbase / Stagehand v4 / MongoDB / Slack: Examples of the surrounding tooling and infrastructure ecosystem he has highlighted in practical agent deployments.
Newsletter Mentions (82)
“Harrison Chase endorsed the view that agent improvement is harness improvement, with key interventions at the tool boundary: context selection, tool availability, failure recovery, and measurement.”
#8 𝕏 Harrison Chase endorsed the view that agent improvement is harness improvement, with key interventions at the tool boundary: context selection, tool availability, failure recovery, and measurement. He said a loop is being built around deepagents for orchestration logic and LangSmith for measurement.
“Harrison Chase said agent workspaces should be durable, inspectable, and swappable, highlighting MongoDB support for LangChain’s Deep Agents virtual file system.”
Harrison Chase said agent workspaces should be durable, inspectable, and swappable, highlighting MongoDB support for LangChain’s Deep Agents virtual file system. Its BackendProtocol lets agent code use read, write, glob, and grep operations while teams choose their production storage layer.
“Harrison Chase shared that LangSmith’s Messages View helps builders debug agents by replaying conversations and tool calls as the agent experienced them, making traces useful beyond infrastructure teams.”
Harrison Chase shared that LangSmith’s Messages View helps builders debug agents by replaying conversations and tool calls as the agent experienced them, making traces useful beyond infrastructure teams.
“Harrison Chase said model labs will build harnesses for their own models while blocking access to rivals’ harnesses, making a lab-independent harness the only cross-model choice.”
#6 𝕏 Harrison Chase said model labs will build harnesses for their own models while blocking access to rivals’ harnesses, making a lab-independent harness the only cross-model choice. He expressed support for LangChain.
“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.”
Related
An AI company whose Threat Intelligence team published a report on misuse of Claude and related countermeasures. The newsletter highlights evolving malicious-use patterns and defensive responses.
Anthropic’s coding agent. It is relevant to AI PMs as a coding workflow product competing in enterprise and community adoption.
An AI coding tool that introduced Projects, a persistent coordinator-agent workflow. The feature moves teams away from task-by-task chats toward a single long-running thread with subagents.
A company building tools for connecting LLMs to data and documents. Here it is noted for releasing a connector that integrates LlamaParse with ChatGPT.
OpenAI’s coding tool/agent used for software development workflows. It matters for PMs as a replacement or alternative in enterprise coding adoption.
A Slack-connected setup or workspace mentioned as being configured using AsideAI. It is relevant as an example of rapid AI-assisted integration setup.
A protocol for connecting agents to external tools and systems in a standardized way. The newsletter mentions setup instructions that can be pasted into an agent to configure MCP.
A framework company for building LLM apps and agents. In this issue it is mentioned alongside Deep Agents and virtual file system infrastructure.
A Claude-based workflow or assistant used for structured journaling and planning. In this newsletter it transforms captured notes into organized outputs and follow-up actions.
LangChain’s evaluation and observability product for AI systems. The newsletter positions it as the measurement layer for an agent orchestration loop.
A workplace messaging and collaboration platform. In this newsletter it appears as an integration target for AI setup and automation.
Autonomous or semi-autonomous AI systems that can plan and take actions across tools and workflows. This is a core AI PM concept central to product design and evaluation.
An orchestration layer or project for building agent loops and control logic. In this newsletter it is described as part of a loop for orchestration logic alongside LangSmith.
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.
A protocol or capability layer mentioned as part of an open, composable extension philosophy for AI tooling. It is grouped with MCP and Plugins.
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.
An SDK for building Claude-based agents and workflows. It is cited as one of the newer harness-style tools replacing older frameworks.
An agent framework or approach that uses summarization middleware to manage context while preserving work history. It emphasizes offloading tool results and older messages into external storage.
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.
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.
Community middleware example for customizing agent behavior and steering tasks in agent frameworks.
Stay updated on Harrison Chase
Get curated AI PM insights delivered daily — covering this and 1,000+ other sources.
Subscribe Free