Using Src to understand your engineering metrics

Last updated: April 6, 2026

Note: this feature is still in beta, please reach out to your Span contact if you'd like this enabled in your Span account.

Overview

Span Src ("source") is an AI-powered assistant embedded in the Span platform that helps you understand your codebase and engineering metrics. It combines deep code intelligence with Span's engineering analytics so you can ask plain-language questions and get accurate, evidence-backed answers — across all of your connected tools. 

Key Capabilities

🔍 Codebase Understanding

Src connects to a your codebase, engineering metrics, project management tools, and team context and builds a rich, searchable knowledge graph.

📊 Engineering Metrics & Analytics

Src connects directly to Span's knowledge graph and can answer questions about your engineering data, including:

  • PR cycle times, merge rates, and review patterns

  • Team and individual activity

  • Asset catalog data (people, teams, repositories, PRs, commits)

  • Any metric surfaced within Span 

Modes

You can interact with SpanSource through:

  • Web UI — a chat interface directly within Span

  • Slack — message the SpanSource bot in your connected workspace 

The SpanSource chat interface offers three modes:

Mode

When to use

Normal

General questions — quick lookups, code explanations, metric summaries

Research

Deep investigations requiring multi-step analysis, cross-referencing multiple sources, root cause analysis

Auto

Lets SpanSource decide which mode best fits your question

In Research mode, SpanSource uses a dedicated research agent that performs iterative, structured investigation — it breaks down problems, uses a task list to track progress, self-reflects on its answers, and validates citations before responding.

File Uploads

You can attach files to your messages. SpanSource will incorporate file contents into the context of its response.

What You Can Ask

About your code:

  • "How does the authentication flow work in the API server?"

  • "Find all places where we handle database connection pooling."

  • "What does the ChatRunService do and how does it relate to the thread lifecycle?"

  • "Show me the directory structure of the payments module."

About your engineering metrics:

  • "What is each team's PR cycle time this quarter?"

  • "Who are the top contributors to the backend repository in the last 30 days?"

  • "Show me PRs with the longest review times."

Cross-cutting investigations:

  • "Investigate where the PR review bottlenecks are for the platform team."

  • "Why might the cycle time have increased last sprint? Check both the code and metrics."

Getting Started

  1. Navigate to SpanSource — open Span and go to the Source section in the left hand navigation.

  2. Start a conversation — type your question in the chat input and press Send. Use the mode buttons to select Normal, Research, or Auto as appropriate.

  3. Use Slack — if your workspace has the Span Slack app installed, you can message SpanSource directly from Slack. 

Tips for Better Results

  • Be specific about the repository or team when asking about metrics: "Show me the PR cycle time for the backend team last quarter" will get better results than "what's the cycle time?"

  • Use Research mode for investigations that span multiple questions or require looking at both code and metrics together.

  • Reference specific files or modules if you know them: "In the apiserver/threads/ module, how does run cancellation work?"

  • Ask follow-up questions — SpanSource maintains conversation history within a thread, so you can drill deeper without repeating context.