Skip to content
Credit card required for Libra Logs Cloud. See plans and billing.
Libra Logs

AI Customer Monitoring

See what’s broken, for which customer, and why.

Set up with your coding agent

Set up Libra Logs (https://libra.exla.ai/logs) in this repo.

1. First ask whether I want to connect Libra Logs to my coding tools so they can search my logs and read change pages. Explain that this writes an MCP connection to my own global Claude Code or Cursor configuration. Only if I explicitly agree, use `npx -y libra-logs@latest --yes --mcp`; otherwise use `npx -y libra-logs@latest --yes`.
2. Run the selected command from the repo root. It waits for browser approval, so run it in the background or with a timeout of at least 10 minutes. Show me its sign-in link and code right away, then wait while I approve the selected workspace. After approval, the installer automatically obtains workspace-scoped API keys. Do not ask me to create or paste an API key manually.
3. If it says it skipped the AI step, ask whether Libra Logs may send this repo's source files to a model. Only if I say yes, rerun the selected command with `--ai-consent` added. Keep `--mcp` only if I approved that connection separately.
4. When it finishes, tell me the branch it created, what it changed, whether my first log was confirmed, and whether the requested coding-tool connection succeeded. Do not push or open a PR unless I ask.
Copy setup prompt

Sign in, add a card, paste. API keys are set up when you approve.

Sentry tells you what broke. Libra tells you which customers it broke for, why, and how to fix it. Know before your customers do.

What you get back

Thousands of log lines become one sentence.

  • Which customer, and how many of their people.
  • Why it failed, in words a new hire understands.
  • The fix, with the log lines underneath as proof.

How it works

Follow one customer from a log line to a fix.

Meet Northwind, a fictional Example customer.

  1. Where it starts

    Thousands of log lines. Almost all of them fine.

    Somewhere in there, one customer's CRM updates have been failing since Tuesday. Nothing crashed, so nothing paged, and the failures disappear into the healthy traffic around them.

    Production logs, as they arriveExample
    INFOglobexGET /api/deals 200 38ms
    INFOinitechPOST /api/invoices 201
    ERRORnorthwindhubspot update rejected: software_used
    INFOumbrellaGET /api/inbox 200 21ms
    INFOhoolicalendar sync ok
    WARNglobexsalesforce: record locked
    INFOnorthwindGET /api/contacts 200 44ms
    ERRORnorthwindhubspot update rejected: software_used
    INFOvandelayPOST /api/signup 201
    ERRORinitechinvoice render timed out
    INFOumbrellaGET /api/threads 200 19ms
    ERRORnorthwindhubspot update rejected: software_used

  2. 01 · Spot

    Libra sorts every line by the customer it belongs to.

    Northwind rises to the top: its CRM updates are failing for 3 of its 8 people. The customers who are fine say so, in one word.

    CustomersExample

    Northwind

    CRM updates failing for 3 of 8 people since Tuesday.

    Globex

    Salesforce rejects edits to locked deals.

    Initech

    Invoice PDFs time out on large accounts.

    Umbrella

    No problems.

    Hooli

    No problems.

    Vandelay

    First logs arrived today.

  3. 02 · Understand

    Why it broke, in a sentence you can read.

    HubSpot rejects the update because the field is read-only on their side. The exact log lines sit right under the sentence, so you can check the claim yourself.

    Northwind · CRM updatesExample
    Needs attention3 of 8 people · since Tuesday

    HubSpot won't accept Northwind's contact updates. The software_used field is read-only on their side, so every save that touches it is rejected.

    The log lines behind it

    hubspot.contacts.update status=400 company=northwind
    error="Property software_used is read-only"
    retry 3/3 failed, contact left unchanged

  4. 03 · Fix

    Paste it into your coding agent.

    Copy the finding into Claude Code, Cursor or Codex. It arrives with its log lines, so the agent starts from the cause instead of a guess.

    claude · ~/appExample

    Pasted from Libra Logs

    Northwind: HubSpot rejects contact updates because software_used is read-only. 3 matching log lines attached.

    > find the cause and propose a fix

    syncContact() sends software_used on every update. HubSpot calculates that field, so stop writing it:

    src/crm/field-map.ts

    - software_used: contact.tools,

    + // software_used is calculated by HubSpot. Never write it.

  5. Every morning

    One briefing. Three customers need attention.

    Open the dashboard and the day starts with who needs you and why. Connect Slack and the same briefing lands in a channel.

    Briefing · Wednesday, 8:00Example

    3 customers need attention

    Everyone else is healthy.

    1. 01

      Northwind

      HubSpot rejects contact updates. software_used is read-only on their side.

    2. 02

      Globex

      A Salesforce validation rule locks deals, so edits never save.

    3. 03

      Initech

      Invoice PDFs time out for accounts with very long invoices.

    On your dashboard every morning. In Slack too, if you connect it.

  6. When it breaks

    Tagged in Slack when a customer breaks from their usual.

    Libra learns what normal looks like for each customer. When Northwind drifts far from it, the right person is tagged with the reason and a log line.

    #customer-healthExample

    Libra LogsAPP9:42 AM

    @maya Northwind just broke from its usual. Its CRM updates are failing far more than normal for this account: HubSpot rejects writes to software_used, a read-only field.

    hubspot.contacts.update status=400 company=northwind
    Open Northwind

  7. From your editor

    Ask your coding agent what's broken for Northwind.

    Connect Libra over MCP and your agent reads customer problems, searches logs and opens the evidence itself, with a read-only key.

    claude · ~/appExample

    > what's broken for Northwind?

    ● libra-logs · whats_broken

    One open problem. HubSpot rejects Northwind's contact updates because software_used is read-only on their side.

    It started Tuesday and affects 3 of their 8 people. Want me to trace where we write that field?

Production logs, as they arriveExample
INFOglobexGET /api/deals 200 38ms
INFOinitechPOST /api/invoices 201
ERRORnorthwindhubspot update rejected: software_used
INFOumbrellaGET /api/inbox 200 21ms
INFOhoolicalendar sync ok
WARNglobexsalesforce: record locked
INFOnorthwindGET /api/contacts 200 44ms
ERRORnorthwindhubspot update rejected: software_used
INFOvandelayPOST /api/signup 201
ERRORinitechinvoice render timed out
INFOumbrellaGET /api/threads 200 19ms
ERRORnorthwindhubspot update rejected: software_used

Northwind and every name here are fictional examples.

How it compares

Datadog tells you something broke. Libra tells you who.

Every competitor cell comes from that vendor's own pricing and docs pages.

Which customers are failing

Libra Logs

Yes, grouped by customer

Datadog

Affected users, if RUM is configured

Sentry

Affected-user counts per issue

Axiom

Not documented

PostHog

Via group analytics

Why, in plain English

Libra Logs

Yes, with the log lines behind it

Datadog

AI assistant (Bits AI)

Sentry

AI debugging (Seer)

Axiom

AI assistant (Spotlight)

PostHog

AI assistant (PostHog AI)

Setup

Libra Logs

One prompt in your coding agent

Datadog

Agent or OTLP ingestion

Sentry

SDK per language, or OTLP

Axiom

SDKs, OpenTelemetry, or HTTP API

PostHog

Setup wizard, or OTLP

Price, 10-person team

Libra Logs

From $29/mo. No seat fees.

Datadog

Usage-based: ingest + indexing

Sentry

From $26/mo, billed annually

Axiom

From $25/mo

PostHog

10 GB free, then per GB

At 100 GB/mo: Libra Pro is $199 flat. Datadog's own published rate comes to ≈$495.71, 15-day retention only.

Where the others are ahead
  • Datadog: Breadth: one platform for logs, metrics, traces, RUM, security, and infrastructure monitoring. Libra Logs does logs and the change pages built on them, nothing else.
  • Sentry: Error tracking depth: issue grouping, release health, and session replay are mature products Libra Logs does not have.
  • Axiom: Cheap high-volume ingest: $25/mo includes 1 TB/mo of data loading before overage.
  • PostHog: Analytics, session replay, feature flags, and experiments are available alongside logs.

Paying Datadog for logs? We'll halve it.

Send us your latest invoice and we'll show you the same log volume at half the price or less.

Send us your bill
Read from each vendor's own pricing and docs pages, 2026-09-23 through 2026-09-25

Updated September 2026

FAQ

Questions, answered

What is Libra Logs?

Libra Logs is an AI logging service. It reads your production logs, works out which customer each line belongs to, and explains failures in plain English with the exact log lines underneath. When something's wrong, it can hand the fix to your coding agent over MCP.

Is Libra Logs a Datadog or Sentry alternative?

For the question "which customers are failing, and why": yes. Libra Logs is not a full APM, metrics, or RUM platform like Datadog, and it doesn't do issue grouping or session replay like Sentry. See the full, sourced comparison.

How much does Libra Logs cost?

Starter is $29/mo for 25 GB. Pro is $199/mo for 100 GB, then $0.20/GB after that. Every plan covers your whole team with no seat fees, and there's no separate charge to keep your logs. Already paying Datadog or Sentry? Send us your bill and we'll beat it.

How do I set up Libra Logs?

Paste one prompt into Claude Code, Cursor, or Codex, or run `npx libra-logs` yourself. It detects your stack, installs non-interactively, and asks before it does anything sensitive: nothing is pushed to your repo, and no source file reaches a model, until you say yes.

Does it work with OpenTelemetry and my existing logs?

Yes. If you already export OpenTelemetry, the installer reuses that setup. If you don't, it detects your stack (Node, Next.js, Python FastAPI, Go) and wires up logging directly, shipping alongside wherever your logs go today.

Can my coding agent use Libra Logs?

Yes. Libra Logs runs an MCP server with tools like `search_logs` and `whats_broken`, so Claude Code, Cursor, Codex, or any MCP-compatible agent can look up what's failing, read the log lines, and start a fix, using a read-only key you can revoke anytime.

What does Libra Logs do with my data?

Your log lines, trace ids, and deploy markers are stored to power search, change pages, and answers. A model (via OpenAI) writes summaries and answers questions from log excerpts; your source files are only sent to a model during setup if you explicitly approve it. Libra Logs never sells your data and never uses your logs or code to train AI models.

Self-serve setup in one prompt: no sales call, no demo.

Setup

Set up in one prompt.

Paste it into Claude Code, Cursor or Codex. Your agent runs the installer, you approve it in the browser, and nothing is pushed to your repo until you say yes.

Paste into Claude Code, Cursor, or CodexCopy prompt
Set up Libra Logs (https://libra.exla.ai/logs) in this repo.

1. First ask whether I want to connect Libra Logs to my coding tools so they can search my logs and read change pages. Explain that this writes an MCP connection to my own global Claude Code or Cursor configuration. Only if I explicitly agree, use `npx -y libra-logs@latest --yes --mcp`; otherwise use `npx -y libra-logs@latest --yes`.
2. Run the selected command from the repo root. It waits for browser approval, so run it in the background or with a timeout of at least 10 minutes. Show me its sign-in link and code right away, then wait while I approve the selected workspace. After approval, the installer automatically obtains workspace-scoped API keys. Do not ask me to create or paste an API key manually.
3. If it says it skipped the AI step, ask whether Libra Logs may send this repo's source files to a model. Only if I say yes, rerun the selected command with `--ai-consent` added. Keep `--mcp` only if I approved that connection separately.
4. When it finishes, tell me the branch it created, what it changed, whether my first log was confirmed, and whether the requested coding-tool connection succeeded. Do not push or open a PR unless I ask.
Or run it yourself: npx libra-logs
Read the transcript

Libra Logs by Exla. Your logs know who's failing, and why. Now you do. Your customer updated their CRM. Some fields never saved. App logs and OTel, user and account context, and deploy and change context come together in Libra: customer activity, failed CRM fields, and evidence to check. See where things go wrong. Daily briefing: @Michael, heads up, some CRM fields aren't saving. HubSpot read-only fields. Salesforce locked records. Ask Libra follow-ups right in Slack. Use it directly through your coding agent. Spot. Understand. Fix. The AI logging service. Libra Logs. Built to understand your customers. Try for free. Self-serve setup in one prompt. exla.ai. The soundtrack is music and sound effects, with no spoken narration.

The context diagram and Slack-style card are designed examples of supported workflows, with customer identities omitted. The terminal is a real Claude Code session connected to a customer workspace through MCP. No raw logs or customer identities are shown.

Same logs. Half the bill.

Same logs, half the bill, and Libra tells you which customer each failure hit, why, and how to fix it.

Paying Datadog or Sentry? Send us your bill and we'll beat it.