🩺 Apex Doctor

Find out why your Apex transaction is slow or broken — in seconds, without leaving VS Code.

Apex Doctor turns raw Salesforce debug logs into a CPU profiler, a query-plan checker, and an AI pair-debugger. Free and open source.

No org? No problem. After installing, run Apex Doctor: Try with Sample Log from the Command Palette to see a full analysis in 10 seconds.

Why Apex Doctor

🔥 A CPU profiler that shows where the time went

Most tools tell you a method took 1,200 ms. Apex Doctor separates self time from time spent in child calls, walks the hot path, and calls out the single bottleneck:

AccountHandler.processAccounts   1,200 ms total ·  45 ms self
  ↳ ContractValidator.validate     890 ms total ·  12 ms self
      ↳ SOQL: Contract query       878 ms self   ← bottleneck

A matching heap profiler attributes memory allocations to the method that made them and warns when you approach the heap limit.

🔎 SOQL Query Plan on every query

Every SOQL row has a Plan button. One click runs Salesforce's Query Plan tool and shows whether the query is selective, which index it uses, its relative cost, and every alternative plan Salesforce considered. No more pasting queries into Developer Console one at a time.

🤖 AI root-cause analysis — even inside your org

One click explains what broke, where, how to fix it, and how to prevent it, with Apex code. Keep asking follow-up questions, or use Ask the Log for natural-language queries like "SOQL that returned more than 500 rows" — matching rows come straight from the parsed log, so nothing is invented.

Works with OpenRouter, Google Gemini, Anthropic, OpenAI, or 🛡️ Salesforce Einstein, which keeps every prompt inside your org's Trust Layer.

🔧 Suggest Fix — preview, then apply

Every issue has a Suggest fix button. Apex Doctor finds the source .cls, applies a deterministic transform where one exists (SOQL-in-loop → bulkified query + Map lookup, missing LIMIT), falls back to AI otherwise, and shows the change in VS Code's diff view. Nothing is written until you confirm.

🔴 Live streaming + Trace Flag Manager

Turn on debug logging for any user straight from VS Code — no Setup tab — then watch their logs arrive in real time. Click any row for a full analysis.

Quick start

1. Install Apex Doctor (links below). The Getting Started guide opens automatically — reopen it any time with Apex Doctor: Open Getting Started Guide.

2. Try it instantly: run Apex Doctor: Try with Sample Log.

3. Connect your org (needed for fetching, streaming, trace flags, and query plans):

   npm install --global @salesforce/cli
   sf org login web --set-default
   # sandbox:
   sf org login web --instance-url https://yourdomain--sandbox.sandbox.my.salesforce.com --set-default
   

4. Analyse real logs: right-click any open log → Analyse this Apex Log, or run Fetch Log from Salesforce.

Install

• VS Code: Visual Studio Marketplace, or search "Apex Doctor" in Extensions
• Cursor, VSCodium, Windsurf, Gitpod: Open VSX
• Offline: download a .vsix from Releases → Extensions → … → Install from VSIX…

Everything else it does

Analysis

• 💡 Plain-English Performance Insights — "62% of runtime is SOQL", "SOQL-in-loop detected"
• 🛑 Issue detection — fatal errors, exceptions, SOQL-in-loop, large and slow queries, slow methods, governor-limit risk, queries on objects you flag
• 📊 Governor limits as colored progress bars (green / amber / red)
• 🔗 Clickable stack traces and inline red squiggles in the log (Problems pane, F8 to step through)
• 🧪 Apex test results panel
• 🎚️ Debug-level recommendations — which categories to raise or lower

Execution structure

• 📈 Activity timeline (SOQL / DML / methods / callouts over time)
• 🗺️ Order-of-Execution map reconstructed from your real transaction
• 🪝 Trigger order per sObject and DML phase, with recursion detection
• 🌊 Flow / Process Builder element timeline, including "element in loop" detection
• ⚡ Async tracer linking @future, Queueable, Batch, and Schedulable logs to their parent
• 🧵 User-journey stitching — the several logs one click produces, reassembled into one action

Compare & history

• 🔀 Compare two logs — summary deltas plus an event-by-event execution-path diff
• 🗂️ Recent analyses (last 10 per workspace) and 🔁 recurring patterns across them

In the editor

• Jump from any method to its .cls line, retrieving the class from the org if needed
• 🛠️ Lightbulb quick-fixes on .cls files
• 🧪 Test coverage overlay in the gutter, with per-class % in the status bar
• ▶️ Anonymous Apex playground — write, run, and analyse in one click

Everything lives under the 🩺 Apex Doctor icon in the Activity Bar: Current Analysis, Recent Logs, and Recurring Issues.

AI providers

Set apexDoctor.provider, then run Apex Doctor: Set LLM API Key. If a configured free model is retired or has no quota on your key, Apex Doctor picks a currently available one for the session and tells you which.

Einstein setup: create an External Client App with OAuth, the sfap_api, api and refresh_token scopes, and the client-credentials flow with a run-as user. Then set apexDoctor.provider to einstein, apexDoctor.einsteinDomain to your My Domain host (e.g. mycompany.my.salesforce.com), and apexDoctor.einsteinConsumerKey to the consumer key, and store the consumer secret via Set LLM API Key. Requires an org entitled for Einstein generative AI.

Commands

Settings

Both tables are generated from package.json on every release.

Privacy

• API keys live in VS Code's encrypted SecretStorage, never in plain-text settings.
• AI features receive a distilled summary (issues, top SOQL, slowest methods, limit metrics) — never the raw log.
• Choose Einstein to keep AI processing inside Salesforce.
• All deterministic analysis — insights, profilers, issue detection, limit parsing — runs locally with no network calls.
• Salesforce access goes through your local Salesforce CLI; Apex Doctor never handles your credentials.

Changelog

See CHANGELOG.md.

Feedback

Found a bug or want a feature? Open an issue — a snippet of the relevant log helps a lot. If Apex Doctor saves you time, a rating on the Marketplace helps other Salesforce developers find it.

License

MIT — see LICENSE.