--- schema: 1 type: video title: 'ACP: The Universal Remote Control for AI Agents — Alex Hancock, Block' description: >- Alex Hancock built one of the clients in this talk the night before he gave it, which is roughly the point. Hancock is a software engineer at Block, works on the open source agent harness Goose, now donated to the Linux Foundation, and maintains the Rust SDK for MCP. His argument is that the agentic stack already has a good standard for agents reaching outward to do things, which is MCP, and that the strongest thing about MCP is not its design but the fact that everyone uses it. What is missing is the other direction, a standard way for client software to tell a harness what to work on and to get updates back. Most harnesses today expose a bespoke interface, and in the worst case exactly one application can drive them. He compares it to needing a different browser for every website. His candidate is ACP, the agent client protocol, which came out of the Zed and JetBrains teams wanting one high quality editor client capable of driving any harness. It runs on JSON RPC and carries sessions, user messages, tool call notifications, and permission requests. It is deliberately extensible through underscore prefixed custom methods, so common patterns can surface from real usage across projects and then move onto a standards track. Hancock demonstrates Zed and a terminal client from Poolside driving the same Goose agent, then the remote HTTP and websocket transport his team specified, since the protocol had none. With remote transports for the protocol, for MCP, and for models, all four pieces of the stack, client, harness, tools, and model, become independently placeable. He expects a real client market to push user experience upward. Speaker info: - https://x.com/alexjhancock - https://www.linkedin.com/in/alexjhancock/ Timestamps: 0:00 - Goose, MCP, and ACP 1:23 - Every harness has a bespoke interface 2:21 - Standards create ecosystems 3:02 - ACP, where it came from and how it works 6:17 - Two clients driving the same agent 7:24 - Remote transport and four movable pieces 9:26 - Why a client ecosystem raises quality --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/YkNulwcc5jk-youtube-002b0bbc7da78d28a5b3 Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- canonical: 'https://liquidgold.cc/watch/YkNulwcc5jk' markdown: 'https://liquidgold.cc/watch/YkNulwcc5jk.md' llms: 'https://liquidgold.cc/watch/YkNulwcc5jk/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T04:05:16.044Z' language: null children: [] videoId: YkNulwcc5jk durationSeconds: 661 publishedAt: '2026-09-09T16:00:03.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 100 endSeconds: 144 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=100' heldReason: rights - ordinal: 1 text: null startSeconds: 144 endSeconds: 156 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=144' heldReason: rights - ordinal: 2 text: null startSeconds: 156 endSeconds: 188 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=156' heldReason: rights - ordinal: 3 text: null startSeconds: 188 endSeconds: 199 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=188' heldReason: rights - ordinal: 4 text: null startSeconds: 199 endSeconds: 379 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=199' heldReason: rights - ordinal: 5 text: null startSeconds: 199 endSeconds: 379 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=199' heldReason: rights - ordinal: 6 text: null startSeconds: 199 endSeconds: 379 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=199' heldReason: rights - ordinal: 7 text: null startSeconds: 390 endSeconds: 445 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=390' heldReason: rights - ordinal: 8 text: null startSeconds: 445 endSeconds: 456 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=445' heldReason: rights - ordinal: 9 text: null startSeconds: 456 endSeconds: 576 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=456' heldReason: rights - ordinal: 10 text: null startSeconds: 456 endSeconds: 576 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=456' heldReason: rights - ordinal: 11 text: null startSeconds: 576 endSeconds: 586 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=576' heldReason: rights - ordinal: 12 text: null startSeconds: 586 endSeconds: 661 watch: 'https://liquidgold.cc/watch/YkNulwcc5jk?t=586' heldReason: rights --- # ACP: The Universal Remote Control for AI Agents — Alex Hancock, Block Alex Hancock built one of the clients in this talk the night before he gave it, which is roughly the point. Hancock is a software engineer at Block, works on the open source agent harness Goose, now donated to the Linux Foundation, and maintains the Rust SDK for MCP. His argument is that the agentic stack already has a good standard for agents reaching outward to do things, which is MCP, and that the strongest thing about MCP is not its design but the fact that everyone uses it. What is missing is the other direction, a standard way for client software to tell a harness what to work on and to get updates back. Most harnesses today expose a bespoke interface, and in the worst case exactly one application can drive them. He compares it to needing a different browser for every website. His candidate is ACP, the agent client protocol, which came out of the Zed and JetBrains teams wanting one high quality editor client capable of driving any harness. It runs on JSON RPC and carries sessions, user messages, tool call notifications, and permission requests. It is deliberately extensible through underscore prefixed custom methods, so common patterns can surface from real usage across projects and then move onto a standards track. Hancock demonstrates Zed and a terminal client from Poolside driving the same Goose agent, then the remote HTTP and websocket transport his team specified, since the protocol had none. With remote transports for the protocol, for MCP, and for models, all four pieces of the stack, client, harness, tools, and model, become independently placeable. He expects a real client market to push user experience upward. Speaker info: - https://x.com/alexjhancock - https://www.linkedin.com/in/alexjhancock/ Timestamps: 0:00 - Goose, MCP, and ACP 1:23 - Every harness has a bespoke interface 2:21 - Standards create ecosystems 3:02 - ACP, where it came from and how it works 6:17 - Two clients driving the same agent 7:24 - Remote transport and four movable pieces 9:26 - Why a client ecosystem raises quality --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/YkNulwcc5jk-youtube-002b0bbc7da78d28a5b3 Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- Canonical: https://liquidgold.cc/watch/YkNulwcc5jk · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.