Feature · Agent context

Failures that arrive already explained.

Microfilm’s agent context is a native MCP server that gives AI coding agents structured test-failure context — the requirement, the expected result, what the tester observed, and a link to the signed capture — instead of a screenshot they can’t parse.

When a test fails, the linked Jira or Linear ticket gets a structured comment — requirement, acceptance criteria, what the tester saw, and a deep link to the capture. Your AI coding agents read the same context through a native MCP server, so the fix loop starts with the full story in hand.

Availability · Team and up — the MCP server and ticket integrations are cloud features.

What Microfilm is · Microfilm is a desktop app and cloud service that records software QA testing as it happens and turns it into signed, tamper-evident, audit-ready evidence — readable by both human auditors and AI coding agents.

MICROFILM Context · TKT-1417
Handed to the agent
Failed
TC-418 · fiscal year
Expected
422 when out of range
Observed
“picker maxed at 2024”
Evidence
Deep link · signed
Read via
MCP server

The problem

A bug that just says "it’s broken" wastes everyone’s time.

A failed test that lands as a one-line ticket kicks off a round of questions: what exactly was tested, what was expected, what did you actually see, where’s the proof? Every hop costs time — and an AI coding agent with no context guesses, which costs more.

How it works

What this gives you.

01

Structured failures, posted automatically

A failed check posts a structured comment to the linked Jira or Linear ticket: the requirement, the acceptance criteria, what the tester observed, and a deep link to the capture. Nobody opens a thread to ask what happened.

02

A native MCP server for coding agents

Claude Code, Cursor, or whatever you run reads the same structured evidence directly through a Model Context Protocol server — not a screenshot it can’t parse, but the requirement, the expected result, and what was actually seen.

03

One source of truth for humans and agents

The tester, the developer, and the agent all work from the same signed record. There’s no separate "agent version" of the truth to drift out of sync.

04

The fix loop starts informed

Because the agent begins with the full context — what was tested, what was expected, what failed, and the evidence — its first attempt is grounded in what actually happened rather than a guess.

Where it fits

Built for testing software in the AI era

Agent-ready context is the hand-off step of the loop. It draws on the signed records from capture (session capture, signed evidence) and shares the structured links that also drive audit export — so the same evidence serves the auditor and the agent.

FAQ

Common questions.

What is the MCP server?

A Model Context Protocol server that lets AI coding agents read Microfilm’s structured test evidence directly — the requirement, acceptance criteria, expected result, what the tester observed, and a link to the capture — instead of parsing a screenshot or a prose ticket.

Which AI coding agents work with it?

Any agent that speaks the Model Context Protocol, such as Claude Code or Cursor. They read the same structured context a human developer sees on the ticket.

Which issue trackers does Microfilm post failures to?

Failures route to the linked Jira or Linear ticket as a structured comment with evidence; Azure DevOps is also supported for syncing requirements and test cases.

Is the agent context server on the free plan?

No. The MCP server and ticket integrations are cloud capabilities on the Team plan and up.

Keep exploring

Start building your evidence trail.

Create a workspace for your team, or download the free capture app and record your first session.