Datadog are doing very cool stuff in this space, worth a follow!
brandSteady
Datadog on X
The largest group on the page is praise, 12 of 53 people, many of them wiring Datadog alerts and traces into agent workflows; the two complaints raised as often as each other are cost and sensitive data ending up in Datadog.
First written 27 Oct · what changed · from 64 posts by 53 people
Tracking posts that say “datadog”
Track this for your marketPosts, last 24h
40
0%vs. previous 24h
Posts, last 7 days
245
118 in the last 72h
Velocity
35/day
7-day average
Momentum
+14%
Last 24h vs. a normal day this week
Posts per day, last 30 days
as of Nov 3, 09:00 UTCWhat changed
as of 27 Oct · checked once a monthFirst written 27 Oct. When what people say changes, it shows here.
Tone· 29 posts from 21 people· written 27 Oct
How do people feel about Datadog?
Praise leads reception: 12 of the 21 people here recommend Datadog, call it the industry leader or say it works for them. The next biggest group, 6 people, is not criticising it either — they describe wiring Datadog alerts, logs and traces into agent workflows as the step that catches real problems.
The praise is short and unreserved rather than argued: people say the security-and-efficiency combination is where Datadog leads the industry, name it flatly in reply to a question about what to use, and report from an account meeting that interesting AI features are coming. 4 people single out the AI work — Bits AI actively helping, evals classifying failure cases cleanly, AI-assisted performance investigation alongside another tool — and one says feeding information into notebooks lets Bits act as if it had internet access.
6 people use Datadog as both trigger and last check in agent workflows: a monitor fires, a workflow wakes an agent that pulls the error trace and opens a fix, and after each release the service is watched in Datadog for new errors. One says that watching step is where real bugs show up, and another has a bot read every alert and summarise the logs before they do. A smaller group, 2 people, likes specific pieces: registering related information in the service catalog, and the log pipeline preview that compares before and after a format change.
- Positive about Datadog13
- Values alerts, traces and agent workflows9
- Praises Datadog's AI features5
- Likes specific Datadog features2
See 5 of 29 postsHide the posts
Automatize parte da sua sustentação Você consegue usar o que você já tem em casa. Na empresa, fiz o seguinte: - usamos assinatura do Claude - criei uma rotina com ambiente próprio, instruções, skills, plugin e modelo - criei um monitor no Datadog para monitorar regras de erro - criei um workflow no Datadog acionado pelo monitor para enviar alerta no Slack e triggar meu agente no Claude - quando dá erro 5xx N vezes em menos de 1h, recebo o alerta no Slack e meu agente é despertado para fazer o trace de erro no Datadog - agente no Claude faz o trace, faz o fix e abre PR completa e humanizada para homologação - eu dou merge, testo e subo para produção Nessa, já vi um alerta durante o final de semana. Amanhã vou ver a PR, testar e subir a correção
Anthropic's Claude Code team says AI now writes about 90% of its code, and reverts fell over 40%. The model isn't the reason. The review layers are. This weekend, on a wealth advisory platform we build, coding agents worked through the issue backlog in batches. Every batch took the same path: Merged. Validated on the dev portal. Released to production. Watched in Datadog for new errors. "Our team uses AI agents" tells a buyer nothing. Ask what happens between the agent's PR and the customer.
$DDOG revenue went from $604M to $3.97B in 5 years that’s 6.5x 📈 gross margin ~80% estimates have revenue at $6.76B by 2028 every AI app needs monitoring, and Datadog is the one most teams default to the growth line here barely blinks
うちも Datadog の Bits AI にバリバリ助けられてます #hcmpl26
5 of the 29 posts behind this answer, one per person. The other 24, and every new one as it’s posted, come with tracking. Get early access
Pain points· 21 posts from 20 people· written 27 Oct
What do people complain about with Datadog?
The complaints divide evenly: 5 people say it is too expensive and 5 worry that sensitive data — prompts, tokens, telemetry — ends up in Datadog. Next, 4 people find the interface and log and trace debugging clunky and error prone.
On price, the specifics are overage and surprise: one team committed to fewer hosts than it runs, with every host above the commitment billed on demand. Another is happy with the choice but says it "can get expensive", counting the cost of no observability too.
On exposure, 5 people raise it: a bug-bounty report of a live password reset token reaching Datadog RUM, a warning that logging raw prompts to an open Datadog instance defeats other protections, and concern that Claude, Datadog and Sentry telemetry can link a local machine to a VPS. One argues LLM tracing is riskier than APM because traces carry system prompts and customer data.
4 people call the interface annoying and log and trace gathering "tedious and error prone", one preferring an agent over "um trace bizarro no datadog". 4 people name limits: vulnerability scans too slow, host data needing SSH access, no compressed-pointer support, Bits AI needing well-fed context. 2 people cover the Windows Agent accepting chains built on compromised eMudhra keys until 7.84.6.
- Too expensive5
- Sensitive data leaks into Datadog5
- Clunky UI and painful logs and traces4
- Product limitations4
- Agent accepted compromised certificates3
See 5 of 21 postsHide the posts
A Datadog contract for 20 hosts. The team runs 31. Every host above the commitment is billed on demand: $660 of a $1,480 bill. Split by charge type. Renegotiate, or bring the hosts back. 👉 link #buildinpublic
Agents love metrics and telemetry over MCP, and the current model is going to break - I love Datadog, but the cost structure is weak against agents that can build dashboards and the increasing ease of deploying ClickHouse and others. They'll need to cannibalize their own pricing model.
评论区有人提醒了一个容易忽略的问题：Claude Code 的遥测数据。 他的思路是，在本地屏蔽 Claude、Datadog、Sentry 相关的遥测流量，尽量避免暴露真实使用环境。 因为即使 API 请求走了 VPS，遥测流量仍可能暴露本地客户端信息，让 Anthropic 有机会将本地环境与 VPS 关联起来。 这么一来，VPS 也未必能起到预期的隔离效果 🫠 所以 Claude 的风控可能远不止 IP，客户端特征、遥测数据都可能是关联线索。 当然，这目前还只是推测，缺乏直接证据。 有人实测过关闭遥测后的效果吗？稳定性有没有改善？ 🤔
the callout on avoiding Kafka and proprietary auth for air-gapping is spot on. LLM observability is fundamentally different from traditional APM (Datadog/Sentry) for 3 physical reasons: 1. Traces Are the Ultimate Toxic Asset: traditional APM captures HTTP routes, query metrics, and error stack traces. LLM tracing captures the raw crown jewels: internal system prompts, retrieved proprietary RAG chunks, multi-turn reasoning chains, and customer PII passed in prompt context. if your observability plane phones home to a multi-tenant cloud, you've created an out-of-band exfiltration path for your entire internal knowledge base. 2. The Operational Burden of Air-Gapped Dependencies: in air-gapped enclaves (defense, banking, sovereign cloud), operational complexity is a security vulnerability. Kafka requires KRaft quorum, JVM memory tuning, and partition rebalancing that turns air-gapped deployments into full-time operations sinks. sticking to Postgres (transactional metadata) + Redis (ephemeral queues) + ClickHouse/S3 (columnar trace storage over MinIO/Ceph) is the sweet spot: completely hermetic, zero external phone-homes, and trivially orchestratable via standard Helm charts. 3. Client-Side Sanitization Before Ingestion: in an air-gapped topology, trace deletion is mathematically brutal (ClickHouse mutations like `ALTER TABLE ... DELETE` rewrite entire compressed columnar parts). having an OSS self-hosted SDK means teams can enforce deterministic client-side PII scrubbing and secret redaction *before* the trace payload ever hits the ingest buffer. building on boring, self-hostable primitives isn't just about avoiding SaaS lock-in; it's the only way enterprise AI survives compliance audits.
I don't know why people are using datadog instead of dash0, the former is one giant mess, not allowing you to check logs in the simple way possible. And I don't even want to mention tracing which is bullshit there.
5 of the 21 posts behind this answer, one per person. The other 16, and every new one as it’s posted, come with tracking. Get early access
Alternatives· 10 posts· written 27 Oct
What are the alternatives to Datadog?
Nothing dominates: 2 people talk about building observability in-house, and beyond that each alternative named — Dynatrace, Chronosphere, Dash0, Kibana, SigNoz, ClickHouse, Basin, a Howdy SIEM — is one person's comparison. Cost and the messiness of logs and traces are what people weigh against it.
The in-house thread is a disagreement, not a trend: one team at Trendyol says it uses no licensed product from a list that includes Datadog, because its SRE and observability teams build their own; another argues cheap code is not the blocker, since the hard part is the knowledge and "being the people to yell at".
Most mentions are single comparisons. Dynatrace is called a great product that "identifies issues better than Datadog". Dash0 is preferred because Datadog's logs and tracing are "one giant mess". Kibana stays only because a company cannot afford to move. SigNoz replaces Datadog on a self-hosted server to save money, and Basin takes gateway logs while Datadog keeps logs and traces. One person weighs Chronosphere as a competitive threat, another wonders whether an in-house Howdy SIEM would have beaten Datadog.
Cost is the thread running through several: one fan of the product says its cost structure is weak against agents that build dashboards and the ease of deploying ClickHouse.
- Building it in-house2
- Dynatrace1
- Chronosphere1
- Dash01
- Howdy SIEM1
- Kibana1
See 5 of 10 postsHide the posts
Trendyol Teknoloji ekibimizde listedeki (vb) hiçbir lisanslı ürünü kullanmıyoruz çünkü SRE ve Observability ekiplerimizin geliştirdiği kendi ürünlerimizi kullanıyoruz: - NewRelic - Datadog - PagerDuty - Incidentio
I don't know why people are using datadog instead of dash0, the former is one giant mess, not allowing you to check logs in the simple way possible. And I don't even want to mention tracing which is bullshit there.
Save $1,700 a year by running these on your own server. Vercel → Dokploy ($240) UptimeRobot → Uptime Kuma ($108) Datadog → SigNoz ($180) GitHub → GitLab CE ($48) Calendly → Cal . com ($120) Airtable → NocoDB ($240) Zapier → n8n ($240) Slack → Mattermost ($87) Dropbox → Nextcloud ($120) Zoom → Jitsi ($170) Confluence → Docmost ($69) Linear → Plane ($120) Entry tiers, one seat each. A 16 GB Netcup VPS runs all twelve for $350 a year.
yup, very similar to what i landed on. currently split between datadog for logs/traces + basin. was originally exploring this exact thing for api gateway request logs separate from regular app otel and went through a few iterations across stacks
Agents love metrics and telemetry over MCP, and the current model is going to break - I love Datadog, but the cost structure is weak against agents that can build dashboards and the increasing ease of deploying ClickHouse and others. They'll need to cannibalize their own pricing model.
5 of the 10 posts behind this answer, one per person. The other 5, and every new one as it’s posted, come with tracking. Get early access
Questions· 6 posts· written 27 Oct
What do people ask about Datadog?
Two questions come up twice each: whether there is an open-source or self-hosted replacement for Datadog, and whether a Datadog PHP client exists. The rest are one-offs — why teams run Datadog and New Relic together, and whether Datadog will move to per-query pricing.
The open-source question is about swapping the product out, not extending it: one person calls Datadog "an immensely powerful database/search engine" and wonders whether something self-hosted could meet their scale, and another, replying to a team that built its own observability, asks whether that Datadog alternative is on GitHub. 2 people ask it.
The PHP question is a discovery, answered by someone else: one person is surprised a Datadog PHP client exists at all and is told "a very popular one" does. 2 people.
The single asks are practical. One wonders why teams use New Relic and Datadog side by side (1 person). Another, replying to a post about agents building dashboards and the ClickHouse threat, asks whether Datadog would move to per-query pricing or whether dashboard-building by agents breaks any model it has (1 person).
- Is there an open-source replacement?2
- Is there a Datadog PHP client?2
- Why run Datadog and New Relic together?1
- Will Datadog move to per-query pricing?1
See 5 of 6 postsHide the posts
Is there an open source/vibecoded way to replace Datadog? It feels like am immensely powerful database/search engine, but I wonder if something self-hosted could meet our scale needs.
Peki kullandığınız kendi ürünleriniz GitHub da mevcut mu . Bazı ürünlerinizin githuba koyduğunuzu biliyorum . Özellikle Datadog un alternatifini merak ediyorum
there's a datadog PHP CLIENT??
a very popular one link
NewRelicとDatadogを併用してるのなんでだろう #hcmpl26
5 of the 6 posts behind this answer, one per person. The other 1, and every new one as it’s posted, come with tracking. Get early access
Wishes· 3 posts· written 27 Oct
What do people want from Datadog?
2 people want Datadog to trace what an agent actually produced, not just the steps it logged; 1 person wants it to stay portable by supporting OpenTelemetry so teams are less locked in.
The agent wish is framed as what would earn payment — "datadog for agents is the one i'd pay for first" — because seeing every step an agent took is what makes it trusted with anything real. The sharper version comes from someone whose lead workflow reported success twice a day for weeks while loading no leads: every step ran, and nothing watched what the steps produced. Their ask is that Datadog for agents count outputs at the destination, rows written and emails delivered, and turn red when the count goes flat, since step logs alone would have shown dozens of green runs.
The portability wish comes from someone whose workplace has already chosen Datadog and is happy with it, but who would favour vendors supporting OTEL to avoid being locked in, noting that switching should get easier and that the relationship matters as much as the tool.
- Better agent observability2
- Stay portable with OpenTelemetry1
See the 3 postsHide the posts
Number 8 is the one I would pay for first. My lead workflow reported success twice a day for 23 days and loaded 0 leads. Every step ran. Nothing watched what the steps produced. Datadog for agents only works if it counts outputs at the destination, rows written and emails delivered, and turns red when the count goes flat. Step logs alone would have shown me 46 green runs.
datadog for agents is the one i'd pay for first, being able to see every step it took is what makes you trust it with anything real
At my work, I don’t have much of a say; they’ve decided to go with Datadog, but I’m happy with that choice. I would favor vendors that support OTEL so that you’re less vendor-locked and have the option to switch if you need to. Now, LLMs make switching providers easier. The relationship and partnership become more important. Datadog has been responsive to our organizational needs, and I don’t have many complaints other than it can get expensive (not having observability is also expensive).
In their words
How do people talk about Datadog?
The phrases people use most, counted once per person.
- 1“logs and traces”2
- 2“datadog for agents”2
- 3“watched in datadog”2
- 4“datadog agent”2
- 5“tls chains”2
- 6“embedded python”2
Posts are shown the way X shows them. One of them yours? Here’s why it’s shown, and how to take it off.
Also in Observability and error monitoring
- Grafana Labs+31%Grafana's biggest theme is 39 people running it in their own stack who say it does the job, ahead of 9 who love it loudly.213 posts / 7d
- Sentry−100%Sentry is talked about warmly: plain praise and enthusiasm for its agent and MCP tooling are level at the top of the reception, while the biggest complaint is that self-hosting is heavy and can cost more than the SaaS.198 posts / 7d
Get every post about Datadog, filtered and labelled
This page is the summary. Tracking gives you the stream behind it for any brand, keyword or @mention: each new pain point, request and question as it’s posted, with the spam and promotions already taken out.