Pinned Tweet
@marcklingen sat down to demo all langfuse features in a single video.
Agent observability, monitoring, evals, prompt management, and what is special about Langfuse.
Great place to start if you are new, or catch up on latest changes.
langfuse.com/watch-demo
Agent observability, monitoring, evals, prompt management, and what is special about Langfuse.
Great place to start if you are new, or catch up on latest changes.
langfuse.com/watch-demo
6
10
29
10,791
Two Langfuse plugins are now available on the @SpaceXSI Marketplace:
→ 𝘓𝘢𝘯𝘨𝘧𝘶𝘴𝘦 𝘖𝘣𝘴𝘦𝘳𝘷𝘢𝘣𝘪𝘭𝘪𝘵𝘺 captures your agents’ model calls, tool executions, and token usage.
→ 𝘓𝘢𝘯𝘨𝘧𝘶𝘴𝘦 gives your agents skills to investigate traces, manage prompts, and run ed by @ClickHouseDB
→ 𝘓𝘢𝘯𝘨𝘧𝘶𝘴𝘦 𝘖𝘣𝘴𝘦𝘳𝘷𝘢𝘣𝘪𝘭𝘪𝘵𝘺 captures your agents’ model calls, tool executions, and token usage.
→ 𝘓𝘢𝘯𝘨𝘧𝘶𝘴𝘦 gives your agents skills to investigate traces, manage prompts, and run ed by @ClickHouseDB
5
1
23
1,145
@SpaceXSI “List this week’s most expensive Cursor turns and what they have in common.”
Ask Cursor, update your instructions, then compare cost, latency, and task completion.
Read the guide: clickhouse.com/blog/cursor-ag…
Ask Cursor, update your instructions, then compare cost, latency, and task completion.
Read the guide: clickhouse.com/blog/cursor-ag…
0
0
3
233
langfuse.com retweeted
Ask Halo challenge
The brief: find out why eligible customer applications were getting rejected.
In seven minutes, @langfuse's Halo:
- found an ignored €6,000 subsidy in traces
- tested a revised prompt and
- made it live, fixing the application workflow
Here's the demo.
The brief: find out why eligible customer applications were getting rejected.
In seven minutes, @langfuse's Halo:
- found an ignored €6,000 subsidy in traces
- tested a revised prompt and
- made it live, fixing the application workflow
Here's the demo.
0
3
12
818
New article on how we built Halo 😇
A couple of decisions:
- Halo uses the same MCP server, skill and docs as external agents accessing Langfuse, so every new Halo functionality is immediately available to everyone.
- The agent loop runs separate from its sandbox.
- We use improve Halo.
Read the full article here: langfuse.com/blog/2026-10-2…
A couple of decisions:
- Halo uses the same MCP server, skill and docs as external agents accessing Langfuse, so every new Halo functionality is immediately available to everyone.
- The agent loop runs separate from its sandbox.
- We use improve Halo.
Read the full article here: langfuse.com/blog/2026-10-2…
15
4
64
7,643
experiments now show which metric moved, and why.
chart quality, cost, and latency across runs. compare each score against your baseline. filter to items under a score threshold to find the regressions, then annotate them from the keyboard.
langfuse.com/changelog/2026…
chart quality, cost, and latency across runs. compare each score against your baseline. filter to items under a score threshold to find the regressions, then annotate them from the keyboard.
langfuse.com/changelog/2026…
2
2
25
1,493
langfuse.com retweeted
Here you see Halo help me get started with evals. It looks at my project, determines that the best first step is capturing user feedback signals, and proposes a list of signals I could capture.
Really helpful when starting out, Halo can be a great sparring partner in getting to tup
Really helpful when starting out, Halo can be a great sparring partner in getting to tup
meet Halo, the assistant that lives inside Langfuse 👀
ask about your traces, sessions and metrics in plain language, or have it build a dataset or dashboard for you. it asks before changing anything.
3
3
20
2,176
meet Halo, the assistant that lives inside Langfuse 👀
ask about your traces, sessions and metrics in plain language, or have it build a dataset or dashboard for you. it asks before changing anything.
ask about your traces, sessions and metrics in plain language, or have it build a dataset or dashboard for you. it asks before changing anything.
14
6
67
8,729
"You need the hold out a part of the dataset to see if the issue was actually fixed. Langfuse allows us to do this with ease."
@traderepublic, a European Neo Bank migrates live LLM workflows to agents with train and exam datasets on self-hosted Langfuse.
langfuse.com/users/trade-re…
@traderepublic, a European Neo Bank migrates live LLM workflows to agents with train and exam datasets on self-hosted Langfuse.
langfuse.com/users/trade-re…
5
3
24
877
How do you make your agent cheaper without it getting worse? @lotte_verheyden's new article walks through it 👇
Wrote about optimizing the cost of your AI agent. This is such an important aspect of building agents successfully, and people often do this "vibes-based". This should offer some structure for anyone going through the process. x.com/i/article/2110…
3
1
20
1,038
Berlin builders: join us, @distil_labs and @runpod on Monday
Bring your 🪢 traces, train your own small model, win €1,000!
x.com/distil_labs/st…
Bring your 🪢 traces, train your own small model, win €1,000!
x.com/distil_labs/st…
We're hosting a Builders Night with @langfuse and @runpod in Berlin next Monday, 19 October, to kick off GTC week.
Post-train, test and deploy your own small language model with our engineers on hand. Pizza, drinks and €1,000 for the winning team.
luma.com/ahiyw2gp
1
1
14
688
Claude agent traces, now in Langfuse.
The Claude Platform exports OpenTelemetry traces natively from Claude Managed Agents, the Messages API and the Claude SDKs, now in public beta.
Inspect turns, tools, latency and cost in Langfuse. Turn real runs into evals to improve agents.
.co/e8G6Ejhk2w
@ClaudeDevs
The Claude Platform exports OpenTelemetry traces natively from Claude Managed Agents, the Messages API and the Claude SDKs, now in public beta.
Inspect turns, tools, latency and cost in Langfuse. Turn real runs into evals to improve agents.
.co/e8G6Ejhk2w
@ClaudeDevs
7
9
53
5,672
second edition of the @langfuse at @aiDotEngineer run club, this time in NYC
great morning loop through Central Park with coffee and breakfast afterwards
thanks everyone for joining!
great morning loop through Central Park with coffee and breakfast afterwards
thanks everyone for joining!
0
0
24
1,867
"The whole pipeline to improve our chat agent happens entirely in Langfuse. Engineers and domain experts work hand-in-hand.."
@DKB_de self-hosts Langfuse to trace and screen its in-app assistant across 10,000 to 20,000 conversations a day. langfuse.com/users/dkb
@DKB_de self-hosts Langfuse to trace and screen its in-app assistant across 10,000 to 20,000 conversations a day. langfuse.com/users/dkb
3
4
21
1,062
how do you know an AI agent is safe to ship at a bank?
@doneyli's answer: golden datasets, domain evaluators, and a PASS/FAIL gate. model risk gets traces and scores, not screenshots.
wed oct 14, 2:55pm, McCarthy stage at @aiDotEngineer NYC
ai.engineer/nyc/2026/sched…
@doneyli's answer: golden datasets, domain evaluators, and a PASS/FAIL gate. model risk gets traces and scores, not screenshots.
wed oct 14, 2:55pm, McCarthy stage at @aiDotEngineer NYC
ai.engineer/nyc/2026/sched…
1
2
25
886
Banks and fintechs now run AI agents across retail, wealth, compliance and ops. Each must be explainable, tested before release and on record for model risk review.
Langfuse for Financial Services, in production at DKB, @traderepublic, @SumUp and @tryramp 🧵
Langfuse for Financial Services, in production at DKB, @traderepublic, @SumUp and @tryramp 🧵
5
1
13
1,932
@traderepublic @SumUp @tryramp Every prompt and model change is tested against golden datasets before release. If it regresses, the release fails.
In production, evals flag policy breaches across every conversation and route them to compliance.
In production, evals flag policy breaches across every conversation and route them to compliance.
1
0
0
209
@traderepublic @SumUp @tryramp Never in the inference path. Self-hosted in your VPC, fully air-gapped, or Langfuse Cloud in the EU, US or Japan.
From compliance monitoring to credit underwriting and customer support.
langfuse.com/financial-serv…
From compliance monitoring to credit underwriting and customer support.
langfuse.com/financial-serv…
0
1
1
147