import yourtraces fromanywhere
connect your current platform, or instrument directly.
Your agent is already running.That’s not the same as working
neatlogs finds what’s failing, works out why, and hands you a fix.
trusted by teams at
VWO + AB Tasty
Runway
CFO.ai
Gobblecube
Vorflux
Passionfroot
CxConnect
Ziva.sh
Cipherstash
Swirls
24 / Silver flowing veilVersion 21 motion · Version 13 material
add the contextyour traces are missing
connect customer feedback and business context to agent runs.
catch failuresother tools miss
neatlogs monitors every run, surfaces failures, and groups related cases into a single incident.
05 sep, 10:16:42 am
search
input
Checkout keeps timing out. Please complete order NL-10482 without charging Maya twice.
what happened14 steps · 24.8s
3 detections- checkout_agent.run. workflow
- route_request. http
- triage_agent. agent
- gpt-4.1. model
- customer history ignoredretrieve_recovery_policytool
- decide_resolutionagent
- lookup_customer. tool
- search_knowledge_base × 9. tool. Deduped
poor judgement
I couldn’t complete the checkout. The checkout service returned a 504 after retrieve_recovery_policy timed out.
AllErrorsSlow
- checkout_agent19 errors · 24.8s
- support_triage3 errors · 12.6s
- returns_agent2 errors · 9.8s
- catalog_agent1 error · 7.3s
Agent runNL-10482 · Checkout
247 spans19 errors
Issues›Checkout timeout
Live issue · 218 affected runs
HTTP 504Checkout requests time out
checkout_agent.runrun_284
refund_agentagent
retrieve_recovery_policytool
new context2 sources added
Evaluation NewRefund policy2 review decisions attached
Attached
Slack New# support-escalationsrefund should have been approved
Attached
AmaraSupport lead
incidents›support_agent · timeout
customer history ignored
The checkout agent started failing immediately after the retry policy changed.
new context2 signals added
Slack · # support-escalationsrefund should have been approved
Linear · SUP-284recovery policy confirmed
related incidents3 matches
import yourtraces fromanywhere
connect your current platform, or instrument directly.
add the contextyour traces are missing
connect customer feedback and business context to agent runs.
refund should have been approvedcatch failuresother tools miss
neatlogs monitors every run, surfaces failures, and groups related cases into a single incident.
ship a fix thatstays fixed
neatlogs watches production to make sure it holds.investigation reportsupport agent · 05 Sep
92%
high confidenceevidence and context agree
retry removedcreate_checkoutv2.8
Sarah
Joneverything your agentsneed to keep working
ask @neatlogs anything
do it in plain english, from Slack or the app. get an answer you can verify.
investigate from the terminal
find the root cause through the CLI or MCP. hand your coding agent the exact fix.
catch failures automatically
detect critical events. route them to the right people with impact, severity, and context.
build dashboards by asking
describe what you want to track in plain language. get a view tailored to your needs.
Slack
Linear
Notion
ChrisEngineeringworks everywhere yourteam does
frameworks
- LangChain
- CrewAI
- LlamaIndex
- Vercel AI SDK
- Pydantic AI
- Agno
- DSPy
- Maestra
coding agents
- Claude Code
- Cursor
- Codex
- Copilot
business tools
- Slack
- Linear
- GitHub
- Jira
- Notion
- Discord
- Webhooks
- Asana
SDKs
- Python
- TypeScript
- OpenTelemetry
- REST API
start free.scale when your agents do
bring your own model key at no extra cost. enjoy unlimited users on every paid plan.
paid plans include a free month. if you do not upgrade, your workspace moves to free.free
for trying neatlogs on a real project.
$0forever
$0forever
what’s included
- 100k spans a month · hard cap
- 3 investigations a month
- traces, sessions & replay
- human evals, unlimited
- 5 detectors
- 2 users · 1 project
- 14-day retention
starter
popularfor teams shipping agents to production.
$99/month
$84/month
billed annually
everything in free, plus
- 250k spans a month · then $10 per 100k
- 50 investigations a month, then $25 per 10
- 5000 AI evaluations a month, then $15 per 2500
- cloud agents open a PR from any issue
- 10 trained classifiers · all detector types
- unlimited users · 5 projects
- PII redaction, roles & webhooks
- 30-day retention
pro
for teams running several agents at scale.
$499/month
$424/month
billed annually
everything in starter, plus
- 1M spans a month · then $8 per 100k
- unlimited investigations & evaluations
- 100 trained classifiers
- unlimited detectors, projects & prompts
- audit log & SSO
- 90-day retention
- priority support in a shared Slack channel
enterprise
for security, compliance, and scale.
customannual pricing
customannual pricing
everything in pro, plus
- custom span volume · negotiated rate
- custom retention
- SAML/SCIM & data residency
- SOC 2 / ISO, DPA & security review
- VPC or self-hosted ingest
- dedicated support with an SLA
a span is one ingested step of an agent run (an LLM call, a tool call, or a retriever step). a typical trace is 5–20 spans.
fair question01 / 08
can’t I just hand the traces to Claude?
for one small run, yes.
at production scale, traces are large, repetitive, and mostly scaffolding. more context costs more and can actually make the answer worse.
neatlogs compares the right runs, isolates the critical spans, and sends Claude or Cursor only the failure, likely cause, and evidence.
for one small run, yes.
at production scale, traces are large, repetitive, and mostly scaffolding. more context costs more and can actually make the answer worse.
neatlogs compares the right runs, isolates the critical spans, and sends Claude or Cursor only the failure, likely cause, and evidence.
still unsure?
we’ll look at your stack with you.
Your agent is already running.Find out if it’s working
connect your production traces in minutes. no credit card required.