Slides: The State of MCP observability: Observable.tools — Alex Volkov and Benjamin Eckel, W&B and Dylibso
Source Video
The State of MCP observability: Observable.tools — Alex Volkov and Benjamin Eckel, W&B and Dylibso
Relationship To World's Fair 2026
These slides are extracted from a public AI Engineer YouTube video connected to World's Fair 2026. Speaker-matched clips are supporting context unless later confirmed as exact session recordings; official livestream recordings are day-level/event-level source material.
Related Scheduled Sessions
- No individual scheduled session mapping has been assigned yet; treat this as an event livestream deck.
Extracted Slides
OCR text:
INNOVATIONPARTNER
aws
PLATINUMSPONSORS
Graphite
WWindsurf
MongoDB
daily
augment code
WorkOs
OCR text:
NT ger ; aera
eae CY a Ks eos
Y Wwe
"Y )
, ar
| Weights i a § baz
| World's Fair |
OCR text:
Graphite Microsoft d'sFair aw
Vorld'sFair AlEngineer World'sF AlEngineer World's NEngin
World'sFair AlEngineer
xpander.ai PRESENTEDBY Microsoft ec
OCR text:
AIE
MCPsareblackboxes
*forobservability
aphite Worir oft Word'sFair aws
fsFair ander.ai WordsFar Moasot Word'sFair se04j Microsoft aws smol?
OCR text:
Why do we care about this
ES * Both build tools that need MCP observability
* Care about DevEx
*« Bad developer and user experience is threat to adoption
« CTOs/CISOs - Familiar observability is necessary for production
reer EN Hacerseis
INES) ree | nae
wd Ceikes se) SE hk ES 600 OO
OCR text:
Fiterbyopnan
run_client In31s Inputs 小色
MCPClient.connect_to_server Path
MCPClient.interactive_sess.. self <mcp.client.session.ClientSession object at 0x12a45a1c28d0>
AIE MCPClient.cal_tool VMCPClient.call_ool add calculate_bmi arguments name weighukg ClientTraces 70
MCPClient.demo_all_tools MCPClient.call_tool D height_m W&BWeave-MCPClient/Server 1.75
MCPClient.call_tool MCPClient.call_tool add calculate_bmi Output Path content [email protected] 小
MCPClient.call_tool create_thumbnail MCPtool calls text 42858
orld'sFair ellum Wo g://app isError annotations nut1 false Docs:wandb.me/mcpop
Fair Microsoft aws smol?
OCR text:
https:/lobservable.tools
ObservableTools
AIE Let'smakeMCPolstanspanent and observabl
AVAVAFRI
SEET SEE ERR
'orld'sFair m Wo
Microsoft aws smol?
OCR text:
AIE
OpenTelemetry - primitives
Traces
POST /api/checkout
getCart()
calculateFraudScore()
performCheckout()
renderJson()
db - select cart...
db - update cart...
OCR text:
OpentTelemetry - primitives
Sink - Observability Platforms
Fhoaracos WO %¢ clastic splunk > XQ Sentry
a Nieree a penne
we u Microsoft AWS ec
ree | 2c 2 aw
OCR text:
Two Deployment Scenarios
for MCP enabled agents
ok Beers renee a
\ |. . Po AV aS fai ]
ne ee a Microsoft a> Ssimol
OCR text:
Different Domains
AIE |mcp-agent-client 7chat.turn 7.60s
4chat.model_call
mcp-agent-client 2.53s
mcp-agent-client POST 2.13s
|mcp-agent-client 2tool.fetch mcp-fetch-server fetch.operation 378.03ms 374.52ms
2chat.model_call
-client 5.07s
Norld'sFair daily Wo Mi
SFNTDV Work
World'sFair PHUNTIST Microsoft W brair Microsoft aws smol?
OCR text:
SameDomains
|mcp-agent-client 15chat.turn 22.53s
AIE |mcp-agent-cllent 6chat.modeLcall 2.49s
POST
mcp-agent-client 2.05s
|mcp-agent-client 4tool.fetch 401.04ms
mcp-fetch-server 3fetch.operation 397.04ms
fetch.http_request mcp-fetch-server 363.21ms
fetch.markdown_conversion mcp-fetch-server 32.90ms
Microsoft orld'sFair -agent-cllent chat.model_call 3.00s
World's AEngrer Microsoft aws smol?
OCR text:
Context Propagation
recs cee: thas client es
name tcovilee. nane
arguments toallisr.argunents
_neta {
__traceContext {
traceld spar Q).traceld
spanId spen ().spanid
traceFlags span ().traceFlags
be3. sik eee ee
}
iH
Bae a });
TAD LCe\*) CS
LY
ss w Microsoft @WS ary?
a Saad all ad Pre Ya) [We wee)
Beas 4
OCR text:
Acknowledge prior work
« MCP Spec Github had several proposals and code examples
ES * RFC #246 by Sam Spencer proposes contextPropagation
« Justin from the Steering Committee suggested a community convention
rather than integrating into spec
* Openinference group started adapters
rE iliuwccananeae yen #25 LA Te place where
ee a Tee om
(Ay panienonivaiany
nealing ene a 5 Reo
vero a
me | ro |: A ©
‘awn aCe colon a Ssimol
OCR text:
param usingOTELand_meta ObservableMcPToOLS MCPCLIENTTRACES
AIE
MCPSERVERTRACES
Vorld'sFair Graphite World'sFair Mi
Weights Vorid
World'sFair Microsoft Word'sFa brain Microsoft aws smol?
OCR text:
WeaveOTel exampleaswell
AIE.local.env.example
#OpenTelemetry endpoint （optional, defaults to http://locaLhost:4318) OTLP_ENDPoINT=https://trace.wandb.ai/oteL/v1/traces
#Weights & Biases API Key for Weave tracing （optionaL） #Get your API key from:https://wandb.ai/authorize
WANDB_API_KEY=
#W&B project in format "entity/project"（e.g.,"myteam/myproject") WANDB_PRoJECT_ID=your_entity/your_project
lorld'sFair raphite World'sFair Mic
Weights baz
Worid'sFair Microsoft Word'sFair rain Microsoft aws 5 smol?
OCR text:
withOpus4&Windsurf MagicMcPmoment
AIE
sFair.um Word'sFair Microsc
SENTR Norld'sFai
World'sFair Microsoft World'sFair braintrus Microsoft aws smol?
OCR text:
MagicMcPmoment
withOpus4&Windsurf
AIE
Great! The test is working.Letme wait a moment for the traces to be
sent to Weave and then checkif the inputs/outputs arenow properly
mapped:
Ranterminalcommand
/mcp-otelSsleep10
Auto-run
MCPTool:wandb/query_weave_traces_tool
orld'sFair
um
Wo
Microsoft
aws
smol°
Aladd'Eain
OCR text:
Magic MCP moment Ican see the attributes are being set with inguts.and outputs.
withOpus4&Windsurf level nputs and output fieldsin Weave.Thismight be because Weave expects a specific format.Let me check the Weave documentation again for the exactformat:
MCPToolwandb/query_wandb_support_bot
AIE sent toWeave and then checkif the inputs/outputsare now properly mapped: Great!The testisworking.Letmewaita moment for the tracestobe Ran with these arguments:
Ran terminalcommand
//mcp-otelSsleep10 Auto-run Output
MCPTool:wandb/query_weave_traces_tool "sources":[ "https://github.com/wandb/weave/tree/naster/weave-js
the attributes are being properly set by looking at the fulltrace data: Inotice that the inputs and output fields are stillempty.Letme check if
sFair lum Worid'sFair licroscments: db/query_weave_traces_tool Ah! The support botindicates that Weave expects wondb.inputs.and the code touse the correctprefixes: wondb.outputs.prefixes,notjust inputs.and outputs..Letme update
SENTRY
World'sFair Microsoft World'sFair trus Microsoft aws smol?
OCR text:
mcp.run - Profiles & Tasks
Example:A"WorkAssistant”Profile
AIE Cursor,Claude Desktog.Winds urforyourownAlageet canalmanage GiHubissuessend Slack messagesand update Notion pages-allthrough Slack+ your one proliel Noton WorkAssistant
Inbound LeadRouter
2/13/202510:31:25AM
Instructions Settings Triggers
Editor
This nev contact just submitted their information through our web form,indicating interest in
assign this lead contact to soneone on our tean,please analyze the contact with the goal of
other indicators.
Mic World'sFair Here*s what I suggest doing1 New contact submission: (（firstNase）{（lastNane））<（（email}）>
Microsoft aws smol?
OCR text:
Recap & Call to action
EG * MCP observability is possible today!
* OTel to the rescue with contextPropagation
* mcp.run will support OTel exports
« W&B support both bespoke and via OTel
ne ee Jaity ce ais
\\
yy . .
aS hte h aad it nd Bee S , rns
Sete eta g
mY aws = semac®
iia Pee LU L(ceto) 1 z
OCR text:
Call to Action
Al Engineers:
* Think about observability
« Work on high level SDK adapters like Openinference
* Work on conventions (genai semantic conventions)
Platform Builders:
* Add OTel support
1w RFCs
coos - ery ule talk to us about your ideas about MCP observability
Wl
_* 7 | 4
apatite’ <3 m Microsoft OWS a?
Wiartielle Eaivr
OCR text:
mae / Alen ;
orid'sFair|. - Mdaily [World *~ ::] am Mic
eT |
f | Se
San a ore | rae
Es a °° Lael
| elite karl |
PE caeons | Woit | brain
OCR text:
| eee aE
eee ed ed
) H . - " . a | 0
Reed . PN | World's Fair | _ > 3 Microsof
“ ’ ie :
2 ae 3
Vi ee of _ PP Ita ne aaa
S SENTD vege TT Me Ae a SET d's Fair
| World's Fair |
aa ed Ne | ) ;
Ld Microsoft | World's Fair | sintrust
Slide-Derived Subjects To Review
Subject extraction uses video title, related session titles/descriptions, transcript context, and OCR text when available. OCR is best-effort and should be reviewed against the embedded slide images.