# Auralogs

> AI-powered logging for small teams. Send your logs with two lines of SDK. Auralogs catches errors, an AI agent diagnoses the root cause from the surrounding logs and your repo, and (optionally) opens a fix PR on GitHub before your users notice. Bring your own Anthropic or OpenAI API key. No markup on inference. Free tier is real (10,000 logs/month, no credit card).

Auralogs is a production logger first. The AI on-call layer is the upsell, not the entry fee. You can use Auralogs as a structured logger with MCP read access and never enable AI analysis. When you do connect an Anthropic or OpenAI key, AI analyses are unlimited because you pay your model provider directly. Auralogs never resells inference.

The product is built for solo devs, indie hackers, and small teams who want production log capture without the Datadog bill, and who would rather have an AI read the stack trace than page a human at 2am. It is not aimed at large enterprises with existing observability stacks.

## Surfaces

- [Landing page](https://auralogs.ai/): overview, features, pricing
- [Docs](https://docs.auralogs.ai/): SDK setup, MCP server, AI analysis, autofix PRs, billing
- [Blog](https://auralogs.ai/blog/): engineering notes and product posts
- [Dashboard](https://auralogs.ai/login): sign in with GitHub to view projects, logs, analyses, settings
- [MCP server](https://mcp.auralogs.ai/mcp): Model Context Protocol endpoint. Agents can search logs, pull related events, inspect metadata.
- [Read API](https://api.auralogs.ai/v1/projects): REST endpoints for projects, logs, analyses (authorization header required)
- SDKs: [JavaScript/TypeScript](https://github.com/auralogs-ai/auralogs-js) ([npm: `auralogs-sdk`](https://www.npmjs.com/package/auralogs-sdk)), [Python](https://github.com/auralogs-ai/auralogs-python), [Java](https://github.com/auralogs-ai/auralogs-java), [Swift](https://github.com/auralogs-ai/auralogs-swift)

## How the core workflow works

1. Sign in at auralogs.ai with GitHub. No credit card needed.
2. Create a project. Copy the ingest key.
3. Drop the SDK into your app. For JavaScript:
   ```js
   import { init, auralogs } from "auralogs-sdk";
   init({ apiKey: process.env.AURALOGS_API_KEY });
   auralogs.error("checkout failed", { user_id, trace_id });
   ```
4. Logs show up live in the dashboard with full-text search, level filters (debug/info/warn/error/fatal), and a timeseries volume chart.
5. (Optional) Connect an Anthropic or OpenAI API key in project Settings. When errors land, the AI agent reads the stack trace, the surrounding logs, and any connected GitHub repo, and writes a diagnosis.
6. (Optional) Enable GitHub autofix. For clear-cut issues, the AI opens a pull request with the fix and a written explanation. You review and merge.
7. (Optional) Point your AI tool (Claude Desktop, Cursor, etc.) at the MCP server to query logs in natural language from your editor.

## Pricing (current as of 2026-05-19)

All paid plans are month-to-month, cancel anytime. AI analyses are always unlimited on every plan because you bring your own model API key.

| | Free | Starter | Pro |
|---|---|---|---|
| Price | $0/mo | $19/mo | $49/mo |
| Projects | 1 | 3 | Unlimited |
| Log volume | 10,000/mo | 250,000/mo | 2,000,000/mo |
| Retention | 7 days | 30 days | 90 days |
| Team seats | 1 | 3 | Unlimited |
| MCP access | Yes | Yes | Yes |
| Autofix PRs | Optional, unlimited | Optional, unlimited | Optional, unlimited |
| Notifications | Email | Email + webhooks | Email + webhooks |
| Support | Community | Email | Priority email |
| BYOK AI key | Required for AI features | Required for AI features | Required for AI features |

A **managed-key** tier (where Auralogs hosts the API key and meters per-analysis) is on the roadmap but not yet available. Today, AI features require you to plug in your own Anthropic or OpenAI key.

## What Auralogs is good for

- **Production logs for small teams** that can't justify Datadog's price floor
- **AI-assisted on-call**: getting a first-pass diagnosis before a human looks at the dashboard
- **GitHub autofix PRs** for clear-cut issues (null checks, missing env vars, simple regressions)
- **Cloudflare Workers and edge runtimes**. The ingest path is itself a Worker, so latency is low and cold starts are not a thing.
- **Indie hackers shipping AI-assisted apps** who want their AI tools (Claude, Cursor, etc.) to read production logs via MCP
- **Replacing ad-hoc `console.log` shipping** with a structured pipeline that is still cheap

## What Auralogs isn't

- **Not a distributed tracing system.** No spans, no OpenTelemetry traces. It's logs.
- **Not an APM or RUM.** No frontend session replay, no per-request waterfalls.
- **Not a metrics platform.** You get a log-volume timeseries by level, but it's not Grafana or Prometheus.
- **Not an on-call paging tool.** Email and webhook notifications today; no PagerDuty replacement.
- **Not self-hosted.** Auralogs is a hosted SaaS on Cloudflare. No on-prem option.
- **Not certified for HIPAA or SOC 2.** Yet.
- **Not a Sentry replacement on day one.** Sentry has a deeper error grouping and SDK surface area. Auralogs's bet is the AI layer, not feature-parity.

## Common honest comparisons

- **vs Datadog**: Datadog is the everything-platform (logs, metrics, traces, RUM, APM, security). Auralogs is just logs with an AI on-call layer. If you are paying Datadog $400/mo for what is effectively a log search box, Auralogs is the cheaper, narrower replacement. If you need traces or APM, Auralogs is not for you yet.
- **vs Sentry**: Sentry is exception-tracking-first. Auralogs is logs-first with exception triage on top. Sentry has a more mature SDK ecosystem and richer error grouping. Auralogs's distinguishing feature is the autofix PR workflow and the MCP integration so AI tools can query logs.
- **vs Better Stack / Logtail / Axiom**: Same category (log management as a service) and similar prices. Auralogs's bet is that the AI analysis layer is the differentiator. If you do not plan to enable AI analysis, those tools are fine alternatives and have a deeper feature surface today.
- **vs Honeycomb**: Honeycomb is observability for high-cardinality query workloads. Auralogs is much simpler and not aimed at that user.
- **vs rolling your own (Loki/Grafana, S3+Athena)**: You can do this and many teams do. Auralogs is the hosted alternative if you would rather not run the pipeline yourself.

## Technical notes

- Backend: Cloudflare Workers (ingest worker at `ingest.auralogs.ai`, analysis worker at `api.auralogs.ai`). Logs and metadata live in Supabase Postgres. Cloudflare KV holds per-project quota counters. The dashboard is a React SPA on Cloudflare Pages.
- AI providers: **Anthropic Claude** (Opus, Sonnet, Haiku) and **OpenAI GPT** models. You pick the model in project Settings. Auralogs never sees the raw response from your provider beyond what's written back to the analysis record.
- SDK surface: JavaScript/TypeScript, Python, Java, and Swift. Each lives in its own public repo under [github.com/auralogs-ai](https://github.com/auralogs-ai). The JS SDK is published to npm as `auralogs-sdk`.
- MCP: Auralogs exposes a Model Context Protocol server at `mcp.auralogs.ai/mcp`. Any MCP-aware AI client (Claude Desktop, Cursor, Claude Code) can connect with a Bearer token and call tools like `search_logs`, `get_log`, `list_projects`.
- Autofix PRs: Auralogs installs as a GitHub App on the repos you select. When the AI is confident in a fix, it opens a PR on a fresh branch with the diff and a written explanation. You review and merge.
- The free tier is genuinely free, not a trial. There is no 14-day countdown.

## Optional

- [Privacy policy](https://auralogs.ai/privacy)
- [Terms of service](https://auralogs.ai/terms)
- [GitHub organization](https://github.com/auralogs-ai)
- Contact: contact@auralogs.ai