Logfire is full-stack observability for the agent era: infrastructure monitoring, service maps, APM, logs, metrics, traces, session replay, feature flags, and AI observability in one OpenTelemetry-native product. The agent trace investigator follows every model call, tool call, retry, database query, API request, and log in one trace, with host telemetry correlated to the same investigation. For model-call telemetry, Logfire’s 10M-record free tier covers a workload that lists at about $5,125/month in Datadog Agent Observability with 30-day retention—assuming one Logfire record per LLM span. Beyond the free tier, Team starts at $49 a month and additional records cost $2 per million with one record-based usage meter.
| Feature | Logfire | Datadog |
|---|---|---|
| Platform coverage | Infra, APM, logs, metrics, traces, replay, flags, and AI | Comparable coverage across separately activated products |
| Architecture | One OpenTelemetry-native product | Datadog agents or OTLP intake with platform extensions |
| Pricing Model | Personal: 10M records free; Team: $49/month + $2/M over 10M | $15/infra host + $31/APM host + $350/M LLM spans + $150/M for 30-day retention |
| Host Fees | None | Infrastructure and APM billed per host |
| AI/LLM Support | Agents, applications, and infrastructure in one project | Separately billed Agent Observability product |
| Query workflow | PostgreSQL-compatible SQL across all telemetry | Product-specific syntax; DDSQL for supported data |
| MCP server | SQL queries plus dashboards, alerts, and issues | DDSQL MCP toolset (preview) |
| Budget control | Hard price cap on Team+ plans | Usage across separately priced products |
Tell us what you run, what you pay, and what your team needs to keep. We’ll give you a straight answer about where Logfire fits.
| Scenario | Datadog | Logfire | Savings |
|---|---|---|---|
| 100M LLM spans/month | $50,125/mo | ~$229/mo (90M additional records) | ~99.5% less |
| 500M LLM spans/month | $250,125/mo | ~$1,029/mo (490M additional records) | ~99.6% less |
Calculation scope
Illustrative model-call telemetry only. It compares standalone Datadog Agent Observability on an annual Pro commitment with 30-day retention against Logfire Team. Datadog infrastructure, APM, logs, and other products are neither required nor included.
Basis: USD; US annual list price. Prices checked .
One Datadog-billed LLM span is modeled as one Logfire telemetry record, with 30-day retention on both sides.
$49 $49 (100M − 10M included) ÷ 1M × $2 $180 $160 annual-plan base $160 (100M − 100K included) ÷ 10K × $3.50 $34,965 100M ÷ 10K × $1.50 $15,000 One Datadog-billed LLM span is modeled as one Logfire telemetry record, with 30-day retention on both sides.
$49 $49 (500M − 10M included) ÷ 1M × $2 $980 $160 annual-plan base $160 (500M − 100K included) ÷ 10K × $3.50 $174,965 500M ÷ 10K × $1.50 $75,000 “Covers 10 million logs, spans, and metrics every month.”
The official calculator shows 20 million records and no extra seats totaling $69/month on Team.
“Additional spans: $3.5 per 10K LLM spans.”
The pricing FAQ publishes annual, month-to-month, and on-demand rates plus separate 30-, 60-, and 90-day retention add-ons.
“Use it to reduce ingestion volume and cost.”
Datadog documents client-side trace sampling as a way to reduce billed ingestion.
Run infrastructure monitoring, service maps, APM, logs, metrics, traces, session replay, feature flags, and AI observability in the same OpenTelemetry-native product. Every signal is available to the same team, in the same project, when an incident crosses from the browser to an API, agent, database, or host.
Start with one agent run, then see every model call, tool call, retry, database query, API request, and log in its trace, with host telemetry correlated to the same investigation. Model and agent views surface the pattern; the trace provides the evidence needed to diagnose why a production run failed.
Logfire's free plan includes 10 million telemetry records each month. If each of those records is an LLM inference span, Datadog Agent Observability with 30-day retention lists at $5,125/month before you add APM, host monitoring, logs, or other Datadog products. On Team and Growth, Logfire remains $2 per additional million records and lets you set a hard price cap.
Datadog's Agent Observability is separately packaged and billed. Logfire puts AI, application, and infrastructure telemetry in the same project and uses PostgreSQL-compatible SQL across it. No moving from product to product—or from query syntax to query syntax—to explain a production failure.
Logfire Cloud, Dedicated, and Self-Hosted run the same product. For sensitive workloads, deploy the official Helm chart in your own Kubernetes environment with your own PostgreSQL and object storage, and keep the same instrumentation, views, SQL, and trace-investigation workflow.
For a model-call-only workload that assumes one Logfire record per LLM span within Logfire's 5 KB average payload allowance, 100 million LLM spans with 30-day retention list at $50,125/month in Datadog Agent Observability versus $229/month in Logfire Team. Logfire is ~99.5% less for that workload. Agent Observability is standalone; Datadog infrastructure and APM products are not required or included in either total. At 500 million LLM spans, the comparison is $250,125 versus $1,029, with Logfire ~99.6% less.
It narrows the unit count, not the bill. Datadog Agent Observability bills only LLM spans, meaning calls to a model provider; it captures tool, workflow, agent, embedding, and retrieval spans without billing them. Logfire bills every span as a record, so the same agent run does produce more billable units here. Compare the published allowances. Datadog Agent Observability is free up to 40,000 LLM spans a month, and Pro is $160 a month for up to 100,000. Every Logfire plan includes 10 million records a month, and additional records are $2 per million. An agent run emits several spans for each model call, so count several Logfire records against every span Datadog would bill. Even at ten Logfire records per billed span, the included Logfire allowance still covers the equivalent of a million Datadog-billed LLM spans before anything is charged, which is ten times the Pro tier ceiling. Worked totals and their sources are in the pricing methodology below.
Logfire prices telemetry as one stream: 10 million records are included, then Team and Growth charge $2 per additional million with no host fees and a hard price cap. Datadog charges separately for Infrastructure Monitoring hosts, APM hosts, logs, and Agent Observability LLM spans. A modern application can use several of those products at once, so the bill grows with product coverage as well as telemetry volume.
Yes. Logfire is a full-stack observability platform: infrastructure monitoring, service maps, APM, logs, metrics, traces, session replay, feature flags, and AI observability all work together in the same OpenTelemetry-native product. Datadog may still be a fit when you need a specific Datadog-only integration or compliance capability.
Yes. Logfire Enterprise Self-Hosted runs the same product in your own Kubernetes environment via the official Helm chart. Use your own PostgreSQL and object storage while retaining the same instrumentation, views, SQL queries, and agent-trace investigation workflow.
Both platforms support LLM tracing, evaluation, and human review. Logfire is designed for investigating what happened in production: a single nested trace connects model calls, tool calls, retries, database queries, API requests, logs, and host signals. Model and agent views surface the pattern; the underlying trace explains the failure.
Get started with 10 million free spans, logs, and metrics per month. No credit card required.