Full Agent Transparency Is Here: See Every Tool Call, Diff, and Reasoning Step in Real Time

v1.0.125 — Agent Activity Transparency & Live Explainability Feed

One of the most common questions from teams evaluating autonomous development platforms is some version of: "But what is it actually doing?" It is a fair question. Handing code changes to an AI agent is a significant leap of trust, and that trust has to be earned through visibility, not faith.

Until now, SaaS Factory's pipeline detail view showed high-level status — Discovering features, Implementing, Tests passing. That was enough to follow the broad arc of a pipeline run, but it was not enough to answer the harder questions: Why did the agent choose this approach? What exactly did it write to that file? How did it reason through the compliance check?

With v1.0.125, every one of those questions is answered directly in the dashboard.

What's in the Feed

The updated Pipeline Detail page streams three kinds of structured information as the agent works:

1. Reasoning Highlights

The agent's chain-of-thought is no longer hidden inside a log table. Every significant decision — why a feature was decomposed a particular way, why one implementation strategy was preferred over another, why a step was flagged for review — appears inline as a highlighted reasoning block. You can read the agent's thinking in plain language as it happens.

2. Tool-Call Collapsibles

Every tool invocation appears as a collapsible row: the tool name, a one-line summary, and elapsed time are visible at a glance. Expanding a row shows the full inputs and outputs. If the agent read a file to understand context before writing, you can see exactly which file it read and what it found. If it searched the codebase for a pattern, you can see the query and the results it acted on.

This turns the agent from a black box into something closer to a pair programmer you can look over the shoulder of.

3. Inline Diff Rendering

Code diffs are rendered with syntax highlighting directly in the activity feed, streamed incrementally as the agent writes. No switching to GitHub, no waiting for the pipeline to finish — you can watch the code being produced, file by file, line by line.

Why This Matters for Enterprise Teams

The explainability gap is the single biggest obstacle to enterprise adoption of autonomous deployment tooling. Security teams need to know what is being deployed. Compliance teams need an audit trail. Engineering leads need to trust that the agent's reasoning is sound before they approve a release.

All of that is now available without leaving the SaaS Factory dashboard:

Approve deployments with confidence — review reasoning and diffs in the pipeline detail view before a release goes out.
Debug failures faster — scroll through the tool-call history to find exactly where a pipeline went off track.
Satisfy audit requirements — the full activity feed is persisted via the agentLogs table and remains accessible after a pipeline completes.

Built on Existing Infrastructure

This release required no new back-end services. The agentLogs table and SSE streaming infrastructure were already in place. What v1.0.125 delivers is structured, diff-aware rendering of that data in the live-pipeline-detail.tsx component — turning raw log entries into a navigable, human-readable activity feed.

Log entries are tagged by type (reasoning, tool_call, diff), which lets the renderer apply the right layout automatically without any post-processing.

What's Next

Full agent transparency is a foundation. Future work will build on this feed to support inline approval/rejection of individual agent actions, anomaly highlighting when agent behaviour deviates from expected patterns, and exportable audit reports for compliance workflows.

For now, open any active or completed pipeline in the dashboard and see exactly what your agents have been up to.

Released in v1.0.125. See the changelog for full version history.