Settings

Access your project settings by clicking Settings in the project sidebar.

Most project settings save automatically after you change them. AI provider keys save when you leave the key field. Watch the status text at the bottom of this page for confirmation.

The page includes your project name, API and MCP keys, analysis behavior, AI providers, model routing, repository settings, autofix mode, and notification channels.

API & MCP keys

Use this section to manage ingest keys and read keys. New projects show one plaintext ingest key and one plaintext MCP/API read key in the startup modal; after that, plaintext keys are only shown when you create or regenerate a key.

• Ingest key — starts with aura_; used by SDKs to send logs.
• Read key — starts with aura_read_; scoped to one project and read-only.

Create additional read keys when you want Codex, Claude, Cursor, or another MCP client to investigate logs without granting write access or connecting GitHub. Use read keys with the MCP server and REST API.

Analysis

The Analysis section is split into three subsections.

Analysis behavior

• Enable AI analysis — top-level toggle. When off, logs still ingest but no analyses run.
• Threshold — minimum log level that triggers analysis (warn, error, or fatal only).
• Environments — restrict analysis to specific environment tags (e.g. production only).
• Schedule — enable periodic reviews and pick the interval.
• Dedup window — how long to suppress repeat analyses for the same error.

AI providers

AI providers are optional. Configure one or both providers here if you want auralogs-generated analyses, scheduled reviews, notifications, or GitHub autofix:

• Anthropic API Key — used for Claude models (analysis with Sonnet 4, autofix PRs with Opus 4.6). Keys start with sk-ant-. Stored encrypted and never logged.
• OpenAI API Key — used for GPT models (analysis with GPT-5 mini, autofix PRs with GPT-5). Keys start with sk- and are at least 20 characters. Stored encrypted and never logged.

You only need to configure the provider(s) you want to use. If a key is rejected by the provider (e.g. revoked), the field shows an inline error with the rejection date — re-saving any valid key clears it.

AI provider key fields save when you tab or click out of the field. This avoids saving half-entered keys while you are typing or pasting.

Model routing

Pick which model handles each pipeline independently:

• Analyze logs with — the model that runs on every error and during scheduled reviews. Choose Claude Sonnet 4 or GPT-5 mini.
• Create fix PRs with — the model that reads your source and writes the autofix PR diff. Choose Claude Opus 4.6 or GPT-5.

If you select a model whose provider has no key configured, an inline amber warning appears beneath the select with a link back to the AI providers subsection. The warning is non-blocking — the setting still saves, but those analyses will be marked as skipped until the key is added.

A common setup: start with analysis only, then add an autofix model if you connect GitHub later.

Repository

Connect a GitHub repository only if you want source-aware analysis and autofix PRs:

1. Click Connect Repository.
2. Authorize the auralog-autofix GitHub App for the specific repo. auralogs only sees the repos you explicitly select — not your whole GitHub account.
3. Pick the repo from the list. Once connected, the owner, name, and default branch appear here with a Disconnect button.

Auto-Fix Mode

Visible once a repo is connected. Choose:

• Disabled — no autofix activity.
• Analysis Only — generate analyses but don’t open PRs.
• Auto-Fix — open a PR for every actionable analysis (subject to dedup protection).

Notifications

Configure how you’re notified about new analyses:

• Email — enabled by default. Sends analysis summaries to your account email.
• Webhooks — available on Starter and Pro plans. Enter a URL to receive JSON POST requests.