--- schema: 1 type: video title: >- MCP Tasks (async): Why Aren't Any Agents Supporting Them? — Cornelia Davis, Temporal description: >- You invoke a tool and expect an answer, but real work takes time, and over that time connections drop, networks blip, and processes crash. Cornelia Davis, a distributed systems veteran who wrote the book on cloud native patterns, argues that this is exactly the gap the MCP tasks specification exists to close, and walks through why almost no agents support it yet. A task lets a tool run long, report progress, and pause for human input without losing its place, which means the interaction has to be durable: it survives the client disconnecting and picks up right where it left off. She demonstrates it with an invoice processing flow, a dashboard tracking task state, and a step that waits for a human to submit input before the backend continues, then traces how the spec evolved from V1 to V2. The design she keeps returning to is a stateless core with the harder long running behavior layered on as an extension, RPC requests replaced by the server pushing updates, and life cycle state carefully mapped so clients know what to resume. Her honest takeaway is that just because you can open a long lived stateful connection does not mean you should, and that getting durable long running tasks right is what will finally let agents handle work that does not finish in a single call. Speaker info: - https://x.com/cdavisafc - https://www.linkedin.com/in/corneliadavis/ Timestamps: 0:00 - What MCP tasks are, and why they're hard 1:29 - A distributed systems point of view 2:34 - A first look at a task running 4:03 - What a task actually allows 4:43 - Why long running work breaks 6:02 - Durability across disconnections 7:04 - Demo: invoice processing dashboard 9:10 - Waiting for human input 11:18 - What changed in tasks V1 12:35 - The stateless core 16:37 - Extensions and server pushed updates 20:09 - V2 and what you need to implement --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/s4r6nk5WsZw-mcp-tasks-async-why-the-heck-aren-t-any-agents-supporting-mcp-tasks-async 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/s4r6nk5WsZw' markdown: 'https://liquidgold.cc/watch/s4r6nk5WsZw.md' llms: 'https://liquidgold.cc/watch/s4r6nk5WsZw/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T06:28:52.295Z' language: null children: [] videoId: s4r6nk5WsZw durationSeconds: 1434 publishedAt: '2026-08-02T20:00:06.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 22 endSeconds: 91 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=22' heldReason: rights - ordinal: 1 text: null startSeconds: 102 endSeconds: 164 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=102' heldReason: rights - ordinal: 2 text: null startSeconds: 102 endSeconds: 164 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=102' heldReason: rights - ordinal: 3 text: null startSeconds: 175 endSeconds: 251 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=175' heldReason: rights - ordinal: 4 text: null startSeconds: 175 endSeconds: 251 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=175' heldReason: rights - ordinal: 5 text: null startSeconds: 263 endSeconds: 286 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=263' heldReason: rights - ordinal: 6 text: null startSeconds: 297 endSeconds: 365 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=297' heldReason: rights - ordinal: 7 text: null startSeconds: 365 endSeconds: 376 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=365' heldReason: rights - ordinal: 8 text: null startSeconds: 444 endSeconds: 557 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=444' heldReason: rights - ordinal: 9 text: null startSeconds: 567 endSeconds: 684 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=567' heldReason: rights - ordinal: 10 text: null startSeconds: 696 endSeconds: 764 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=696' heldReason: rights - ordinal: 11 text: null startSeconds: 696 endSeconds: 764 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=696' heldReason: rights - ordinal: 12 text: null startSeconds: 776 endSeconds: 1002 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=776' heldReason: rights - ordinal: 13 text: null startSeconds: 776 endSeconds: 1002 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=776' heldReason: rights - ordinal: 14 text: null startSeconds: 1002 endSeconds: 1014 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=1002' heldReason: rights - ordinal: 15 text: null startSeconds: 1014 endSeconds: 1219 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=1014' heldReason: rights - ordinal: 16 text: null startSeconds: 1014 endSeconds: 1219 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=1014' heldReason: rights - ordinal: 17 text: null startSeconds: 1219 endSeconds: 1230 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=1219' heldReason: rights - ordinal: 18 text: null startSeconds: 1230 endSeconds: 1434 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=1230' heldReason: rights - ordinal: 19 text: null startSeconds: 1230 endSeconds: 1434 watch: 'https://liquidgold.cc/watch/s4r6nk5WsZw?t=1230' heldReason: rights --- # MCP Tasks (async): Why Aren't Any Agents Supporting Them? — Cornelia Davis, Temporal You invoke a tool and expect an answer, but real work takes time, and over that time connections drop, networks blip, and processes crash. Cornelia Davis, a distributed systems veteran who wrote the book on cloud native patterns, argues that this is exactly the gap the MCP tasks specification exists to close, and walks through why almost no agents support it yet. A task lets a tool run long, report progress, and pause for human input without losing its place, which means the interaction has to be durable: it survives the client disconnecting and picks up right where it left off. She demonstrates it with an invoice processing flow, a dashboard tracking task state, and a step that waits for a human to submit input before the backend continues, then traces how the spec evolved from V1 to V2. The design she keeps returning to is a stateless core with the harder long running behavior layered on as an extension, RPC requests replaced by the server pushing updates, and life cycle state carefully mapped so clients know what to resume. Her honest takeaway is that just because you can open a long lived stateful connection does not mean you should, and that getting durable long running tasks right is what will finally let agents handle work that does not finish in a single call. Speaker info: - https://x.com/cdavisafc - https://www.linkedin.com/in/corneliadavis/ Timestamps: 0:00 - What MCP tasks are, and why they're hard 1:29 - A distributed systems point of view 2:34 - A first look at a task running 4:03 - What a task actually allows 4:43 - Why long running work breaks 6:02 - Durability across disconnections 7:04 - Demo: invoice processing dashboard 9:10 - Waiting for human input 11:18 - What changed in tasks V1 12:35 - The stateless core 16:37 - Extensions and server pushed updates 20:09 - V2 and what you need to implement --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/s4r6nk5WsZw-mcp-tasks-async-why-the-heck-aren-t-any-agents-supporting-mcp-tasks-async 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/s4r6nk5WsZw · 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.