Videos ghJmWQCIHRM
The agent-ready web: Simplify user actions with WebMCP — Tara Agyemang, Google
Scene timeline
62 shot(s).
keyframes kept every frame deduplicated
What was stored
- cues
- 206
- whisperx 206
- chunks
- 37
- from 206 cues
- keyframes
- 33
- kept of 62 captured
- frames with text
- 33
- 1,005 lines read
- chapters
- 0
- from the source metadata
- keyframe bytes
- 6.0 MB
- word timings on 206 cues
Provenance
| stage | state | model | started | took |
|---|---|---|---|---|
fetch |
done | — | 2026-08-10 19:22 | 1m 38s |
stt |
done | — | 2026-08-10 19:23 | 21s |
chunk |
done | — | 2026-08-10 19:24 | 0s |
text_embed |
done | — | 2026-08-10 19:56 | 0s |
keyframe |
done | — | 2026-08-10 19:24 | 1m 49s |
ocr |
done | — | 2026-08-10 19:26 | 16s |
frame_embed |
done | — | 2026-08-10 19:56 | 6s |
Frames, and what the machine read
-
- AlEngineer0.98
- EUROPE1.00
-
- PRESENTINGSPONSOR1.00
- Google DeepMind1.00
-
- PLATINUM SPONSORS0.98
- # Braintrust0.96
- WorkOS OpenAI0.95
-
- Google1.00
- The agent-ready web1.00
- Simplify user actions with WebMCP1.00
- Tara Agyemang / April 20260.97
- AlEng0.96
- EUROP1.00
-
- Google1.00
- AIE1.00
- The agent-ready web0.98
- ★0.99
- Simplify user actions with WebMCP1.00
- Tara Agyemang / April 20260.99
- Engineering the future of Al1.00
- AlEngir0.97
- EUROPE1.00
-
- docs.google.com - to exit full screen, press0.97
- AIE1.00
- Google!1.00
- Google1.00
- ★0.99
- Google1.00
- ★1.00
- ★1.00
- ★1.00
- Google1.00
- Braintrust1.00
- WorkOS OpenAI0.95
- AlEng0.92
- EURO0.93
-
- + Aak Oem0.83
- AUsens/taraojo/Projecta/concert-tickets/index.homi0.82
- BEATDROP1.00
- Featured Events About0.97
- Search events, arfists0.96
- ★1.00
- Feel the Live Music0.99
- AIE1.00
- ★1.00
- ★1.00
- Experience1.00
- ★1.00
- ★1.00
- ★1.00
- Find your next unforgettable night with BeatDrop.1.00
- Browse Events1.00
- Hello, Tara0.97
- How can I help you today?0.99
- Discuss a topic with me0.92
- How can I use Gemini Live?0.95
- Do I have any meeting conflicts next week?0.98
- Featured Events1.00
- Sharig "BeatDrop ( COocer Tikkents"0.76
- Type to ask about a tab0.96
- Google1.00
- AlEngineer0.97
- EUROPE1.00
- AlEngir0.91
- EUROR0.90
-
- purchase 2 tickets to the afrobeat festival0.99
- on 28th december, my budget is 200 per0.98
- AIE1.00
- person. pick the best seats, as close the1.00
- the front as possible0.99
- AlEngineer0.96
- EUROPE1.00
- AlEngil0.96
- EUROPE1.00
-
- StaticT0.98
- StaticT1.00
- StaticT1.00
- image "0.94
- <c0.88
- https:/0.99
- Featured Events1.00
- 4b6ad7a0.97
- StaticT1.00
- src="https1.00
- heading1.00
- it=crop&an1.00
- StaticT1.00
- StaticT1.00
- afrobeat0.96
- StaticT0.98
- ★1.00
- image1.00
- AIE1.00
- https:/1.00
- ★1.00
- StaticT1.00
- b30b53o0.92
- ★1.00
- ★1.00
- ★1.00
- heading1.00
- StaticT0.99
- StaticT1.00
- Afrobeat Festival1.00
- Summer Vibes Festival0.98
- StaticT1.00
- heading1.00
- December 28, 20260.96
- 人y12.20260.95
- button1.00
- Elko Atiantk. Lagos0.84
- Central Park, NY0.96
- <C0.78
- button1.00
- button1.00
- £1501.00
- £890.99
- button1.00
- src="https1.00
- image1.00
- it=crop&an1.00
- https:/1.00
- bd47b1o0.99
- StaticT1.00
- heading1.00
- StaticT1.00
- StaticT0.89
- Engineering the future of Al0.99
- AlEngir0.96
- EUROP1.00
-
- ***0.58
- AIE1.00
- WebMCP1.00
- ★1.00
- ★1.00
- Web Model Context Protocol1.00
- Google1.00
- AlEngineer0.96
- EUROPE1.00
- AlEngir0.95
- EURO1.00
-
- WebMCP - Model Context Tool Inspector0.98
- WebMCP Tools0.95
- description1.00
- inputSchema1.00
- name1.00
- Start a new maze game. Generates a fresh1.00
- right exit.1.00
- maze and places the player at the top-left1.00
- corner. The goal is to reach the bottom-0.99
- start_game1.00
- ★1.00
- AIE1.00
- Copy as ScriptToolConfigCopy as JSON1.00
- ★1.00
- ★1.00
- ★1.00
- ★1.00
- MAZE ESCAPE0.99
- Interact with the Page0.98
- User Prompt0.98
- You are trapped in a maze. Ask the AI agent to1.00
- help you escape!1.00
- Start a new maze game.1.00
- The agent can use the start game tool to begin.0.96
- Send1.00
- Reset1.00
- Update Gemini API key0.99
- Copy trace1.00
- Tool1.00
- "start_game"1.00
- Input Arguments1.00
- 0}0.69
- Engineering the future of Al0.99
- AlEngir0.92
- EUR1.00
-
- WebMCP - Model Context Tool Inspector0.97
- WebMCP Tools0.97
- description1.00
- inputSchema1.00
- name1.00
- Start a new maze game. Generates a fresh maze0.99
- and places the player at the top-left corner. The goal0.99
- start_game1.00
- is to reach the bottom-right exit.1.00
- Copy as ScriptToolConfigCopy as JSON0.99
- AIE1.00
- Interact with the Page1.00
- ★1.00
- ★1.00
- ★1.00
- MAZE ESCAPE0.97
- User Prompt1.00
- You are trapped in a maze. Ask the AI agent to0.98
- Start a new maze game.1.00
- help you escape!1.00
- The agest can use the start_game tool to begin.0.95
- Send1.00
- Reset1.00
- Update Gemini API key0.99
- Copy trace1.00
- Tool1.00
- "start_game"1.00
- Input Arguments1.00
- 0}0.82
- Engineering the future of Al0.99
- AlEngir0.93
- EURO1.00
-
- omelabs.github.io/webmcp-tools/demos/webmcp-maze/0.99
- WebMCP - Model Context Tool Inspector0.98
- description1.00
- inputSchema name1.00
- Start a new maze game. Generates a fresh maze0.99
- is to reach the bottom-right exit.0.98
- and places the player at the top-left corner. The goal0.99
- start_game1.00
- Copy as ScriptToolConfig Copy as JSON0.99
- *★★0.67
- ★1.00
- Interact with the Page0.99
- AIE1.00
- ★1.00
- User Prompt1.00
- ★1.00
- Start a new maze game.0.98
- ★1.00
- ★1.00
- ★1.00
- MAZE ESCAPE0.98
- You are trapped in a maze. Ask the AI agent to0.98
- help you escape!1.00
- Send1.00
- Reset1.00
- Update Gemini API key1.00
- Copy trace0.98
- The agent can use the start game tool to begin.0.98
- Tool1.00
- "start_game"1.00
- Input Arguments1.00
- {0}0.67
- Execute Tool0.96
- AlEngineer0.97
- EUROPE1.00
- AlEngi0.95
- EUR0.82
-
- ☆0.94
- MOVES1.00
- 2/1001.00
- HOLDING1.00
- WebMCP - Model Context Tool Inspector0.98
- requareu : L uarecczon0.67
- Pick up a collectible item1.00
- (key or dynamite) from0.99
- only carry one item at a1.00
- your current cell. You can1.00
- pickup1.00
- time. Drop your current0.99
- item first if needed.1.00
- ★1.00
- Interact with the Page0.99
- AIE1.00
- ★1.00
- User Prompt1.00
- ★1.00
- Start a new maze game, then show me what's around, an0.99
- ★1.00
- ★1.00
- ★1.00
- Send1.00
- Reset1.00
- Update Gemini API key0.98
- Copy trace1.00
- User prompt: "Start a new maze game."1.00
- AI calling tool "start_game" with ()0.96
- Tool "start_game" result: {"message":"A new maze has been generated! Use0.99
- 'look' to see your surroundings, then 'move' to navigate.","mazeSize":0.99
- {"rows":10,"cqls":10},"startPosition":{"row":0,"col":0},"exitPosition":0.99
- {"row":9,"col":9}}0.98
- AI result: A new maze has been generated! Use 'look' to see your0.99
- surroundings, then 'move' to navigate. The maze is 10x1e, starting at0.99
- (0,0) and the exit is at (9,9).0.98
- AlEngineer0.97
- EUROPE1.00
- AlEng0.91
- EUR1.00
-
- MOVES1.00
- 2/1001.00
- HOLDING1.00
- WebMCP - Model Context Tool Inspector0.98
- time. Drop your current0.98
- item first if needed.1.00
- Interact with the Page0.98
- User Prompt1.00
- Start a new maze game, then show me what's around, and then move1.00
- north.1.00
- ★1.00
- AIE1.00
- ★1.00
- ★1.00
- Send1.00
- Reset1.00
- Update Gemini API key0.97
- Copy trace1.00
- ★1.00
- ★1.00
- ★1.00
- User prompt: "Start a new maze game."1.00
- AI calling tool "start_game" with {}0.99
- Tool "start_game" result: {"message":"A new maze has been generated! Use0.99
- 'look' to see your surroundings, then 'move' to navigate.","mazeSize":0.99
- {"rows":10,"cols":10},"startPosition":{"row":0,"col":0},"exitPosition":0.99
- {"row":9,"col":9}}1.00
- AI result: A new maze has been generated! Use 'look' to see your0.99
- surroundings, then 'move' to navigate. The maze is 10x1e, starting at0.99
- (0,0) and the exit is at (9,9).0.99
- Tool1.00
- "drop"1.00
- Input Arguments1.00
- AlEngineer0.97
- EUROPE1.00
- AlEngi0.98
- EURC0.89
-
- ☆0.97
- MOVES1.00
- 2/1001.00
- HOLDING0.99
- WebMCP - Model Context Tool Inspector0.97
- { "type": "object",0.97
- Move the player one cell1.00
- "properties": { "direction": {0.99
- Returns success or failure0.98
- in a cardinal direction1.00
- (north, south, east, west).1.00
- "west" ], "description":0.98
- "type": "string", "enum": [0.98
- "north", "south", "east",0.99
- "The1.00
- move1.00
- with a reason.1.00
- direction to move." }},0.98
- "required": [ "direction" ] }0.98
- Pick up a collectible item0.98
- (key or dynamite) from0.99
- *★★0.64
- AIE1.00
- ★1.00
- ★1.00
- your current cell. You can0.97
- only carry one item at a0.97
- time. Drop your current1.00
- pickup1.00
- item first if needed.1.00
- ★1.00
- ★1.00
- ★1.00
- ★1.00
- Interact with the Page0.99
- User Prompt1.00
- Start a new maze game, then show me what's around, and then move1.00
- north.1.00
- Send1.00
- Reset1.00
- Update Gemini API key1.00
- Copy trace1.00
- User prompt: "Start a new maze game."1.00
- AI calling tool "start_game" with {}0.98
- Tool "start_game" result: {"message":"A new maze has been generated! Use0.98
- 'look' to see your surroundings, then 'move' to navigate.","mazeSize":0.99
- {"rows":10,"cols":10},"startPosition":{"row":0,"col":0},"exitPosition":1.00
- {"row":9,"col":9}}0.99
- Engineering the future of Al1.00
- AlEngi0.96
- EURO1.00
-
- MOVES1.00
- EXPLORED 4/1000.98
- HOLDING0.99
- WebMCP - Model Context Tool Inspector0.98
- { "type": "object",0.97
- Move the player one cell1.00
- "properties": { "direction": {0.99
- Returns success or failure1.00
- in a cardinal direction1.00
- (north, south, east, west).0.98
- with a reason.0.99
- "west" i, "description":'"The0.96
- "type": "string", "enum": [0.99
- "north", "south", "east",0.98
- direction to move." }},0.98
- "required": ["direction" ] }0.96
- move0.98
- ★1.00
- Pick up a collectible item0.98
- (kev or dvnamite) fmm0.96
- AIE1.00
- ★1.00
- Interact with the Page1.00
- ★1.00
- ★1.00
- ★1.00
- ★1.00
- User Prompt1.00
- Send1.00
- Reset1.00
- Update Gemini API key1.00
- Copy trace1.00
- User prompt: "move down, then right"0.99
- AI calling tool "move" with {"direction":"south"}1.00
- Tool "move" result: {"success":true,"position":0.99
- {"row":1,"col":0},"atExit":false,"moveCount":1}1.00
- AI calling tool "move" with {"direction":"east"}0.99
- Tool "move" result: {"success":true,"position":0.99
- {"row":1,"col":1},"atExit":false,"moveCount":2}1.00
- AI result: I moved you down, then right. You are now at row 1, column 1.1.00
- Google DeepMind1.00
- AlEngii0.93
- EUROP1.00
-
- MOVES1.00
- 8/1000.99
- HOLDING0.98
- WebMCP - Model Context Tool Inspector0.99
- current inventory, ana1.00
- whether the exit is here.0.98
- { "type": "object",0.98
- Move the player one cell0.98
- "properties": { "direction": {0.99
- (north, south, east, west).1.00
- in a cardinal direction0.99
- "type": "string", "enum": [0.99
- "north", "south", "east",0.99
- move1.00
- with a reason.1.00
- Returns success or failure1.00
- "west" ], "description": "The0.96
- direction to move." }},0.97
- "required": [ "direction" ] }0.96
- ★1.00
- AIE1.00
- (kev nr dvnamite) from0.93
- Pick up a collectible item1.00
- ★1.00
- Interact with the Page0.99
- ★1.00
- ★1.00
- User Prompt1.00
- Send1.00
- Reset1.00
- Update Gemini API key0.99
- Copy trace1.00
- {"row":1,"col":2},"atExit":false,"moveCount":3}1.00
- AI calling tool "move" with {"direction":"north"}0.98
- Tool "move" result: {"success":true,"position":0.99
- {"row":0,"col":2},"atExit":false,"moveCount":4}1.00
- AI calling tool "move" with {"direction":"east"}0.99
- Tool "move" result: {"success":true,"position":0.99
- {"row":0,"col":3},"atExit":false,"moveCount":5}0.99
- Engineering the future of Al0.98
- AlEngi0.98
- EUROP1.00
-
- EXPLORED0.99
- 1/1000.97
- HOLOING GREENKEY0.99
- WebMCP - Model Context Tool Inspector0.99
- current inventory, ana0.99
- whether the exit is here.0.99
- Move the player one cell0.99
- { "type": "object",0.98
- "properties": { "direction": {0.99
- in a cardinal direction1.00
- (north, south, east, west).1.00
- "type": "string", "enum": [0.99
- "north", "south", "east",0.99
- move1.00
- with a reason.1.00
- Returns success or failure1.00
- "west" ], "description": "The0.96
- direction to move."}},0.99
- "required": [ "direction" ] }0.95
- ★1.00
- AIE1.00
- ★1.00
- (kev or rdvnamite) from0.91
- Pick up a collectible item1.00
- ★1.00
- Interact with the Page0.98
- ★1.00
- ★1.00
- ★1.00
- User Prompt1.00
- Send1.00
- Reset1.00
- Update Gemini API key0.98
- Copy trace0.99
- Tool "move" result: {"success":true,"position":1.00
- {"row":0,"col":6},"atExit":false,"moveCount":12}1.00
- AI calling tool "look" with {}0.97
- Tool "look" result: {"position":("row":0,"col":6},"openDirections":0.99
- ["east","west"],"collectibleHere":null,"inventory":"key_green","atExit":f1.00
- - not yet visible through fog of war", "mazeSize":0.98
- {"rows":10,"cols":10},"moveCount":12}0.99
- AlEngineer0.97
- EUROPE1.00
- AlEngi0.96
- EUROP1.00
Transcript
206 cues· 3,183 words· 16,615 chars
- 0:15 Hello?
- 0:16 Hello, can you hear me OK?
- 0:18 OK, cool.
- 0:20 Let's get started.
- 0:22 So we are going to be talking a little bit about WebMCP.
- 0:26 Just out of curiosity, has anybody already played around with WebMCP?
- 0:32 Only a few people.
- 0:32 OK, great.
- 0:33 Those few people, you have a bit of a head start.
- 0:35 But for everyone else, we'll be going into a bit more of the background, how it works, what it does.
- 0:40 So my name is Tara.
- 0:42 I am part of the Google Chrome team.
- 0:45 I'm a developer relations engineer, and I'm here with a few of my colleagues from Google Chrome alongside
- 0:52 the DeepMind team too.
- 0:53 So we'd be really interested to talking to you afterwards around the DeepMind booth if you have thoughts around web and AI and the intersection between the two.
- 1:03 That is where my focus is these days.
- 1:07 So let's get into it.
- 1:09 The, let's say, past few decades, we have been building the web for human actions and human eyes, and we've been trying to optimize for that.
- 1:21 But these days, it's not just humans that are using the web.
- 1:24 We have agents using the web on human behalf, too.
- 1:30 And we are seeing an increasing number of agents using the web.
- 1:35 But the problem is the agents are having to do so much work to do simple actions on the sites that we've built.
- 1:45 And just to give you a bit of an example of this, this is a website that I've live coded.
- 1:51 And it's a concert website for selling tickets for concerts.
- 1:55 And we have Gemini and Chrome panel on the side here.
- 1:59 And let's say you've come along to this website.
- 2:02 and you've typed this prompt, you want to buy two tickets to the Afrobeats Festival, you've given it the details.
- 2:08 The AI agent has to do so much work to make this happen.
- 2:12 So it'll probably look at the HTML because usually the agents will pass the entire DOM just to understand what's happening on your page.
- 2:21 Then it will look into the accessibility tree just to understand the structure of your HTML page.
- 2:26 Then maybe it'll take a screenshot of the page, analyze all the different elements that it couldn't see in the HTML and the accessibility tree.
- 2:34 And then maybe it will measure how far down it needs to click, how far across, where the exact element that it needs to click, and then it'll click that element.
- 2:43 And as you can see, this process is quite long.
- 2:45 It can be brittle, and I don't even want to guess at how many tokens you've probably just used trying to do this.
- 2:51 It's probably a lot.
- 2:52 And then after all that, maybe your ad has loaded at the top of the page, pushed all your content down, and your AI agent couldn't even click the right place in the end.
- 3:02 So there's so much to think about, but before we go into this proposed web standard, it's worth mentioning that you can do so much by improving web foundations first.
- 3:15 So making your site accessible for everyone makes it accessible to AI agents by default.
- 3:23 So if you improve your semantic HTML, if you focus on robust accessibility standards, and if you improve your page performance, make it load really quickly, think about those core web vitals, and then improve really good user experience flows through your site, you're already halfway to getting an agent-ready website.
- 3:47 And it's only once you have those in place that it makes sense to start thinking about WebMCP.
- 3:53 So if you're not already aware, the Web Model Context Protocol is a proposed web standard, and that gives you the ability to define your site's capabilities as structured tools for your AI agents to use.
- 4:10 And so you might have heard references to this as the USB-C of AI agent interactions.
- 4:16 And that's because instead of any agent guessing what your website does, you're kind of giving the AI agent a menu of tools that it can use and actions that it can take.
- 4:31 And so because of this, we're seeing that WebMCP significantly improves the performance and the reliability of agents navigating your website.
- 4:43 So let's see it in action.
- 4:47 Hopefully, Gemini treats me well today.
- 4:50 So this is the maze escape game built by our team in Chrome DevRel.
- 4:55 And just on the side here, we have a Chrome extension.
- 5:00 I'll show you a link to that afterwards.
- 5:02 But this is the model context tool inspector.
loading