Ask the repo. Get a line.
Exact lookups over Robot Framework test repos: one file:line per hit, so people and agents stop reading whole suites.
tests/cases.robot | ticket PROJ-2120
- $ python -m robot_mind --root . ticket PROJ-2120
- test Change Zone NEW Lists Zones
- tests/cases.robot:6
- [bug:PROJ-2120 role:role_team_lead op:GET /api/zones/list]
tests/cases.robot | callers "Open Cases Page"
- test Change Zone NEW Lists Zones tests/cases.robot:9
- test Cases Page Opens For QC tests/cases.robot:13
One file:line per hit. No suite dump.
Agents stop burning tokens on searching.
Characters read to answer one question on a real API test repo of 553 files. Shorter output is cheaper.
robotframework-mindgrep, then open the filescharacters of output, log scale
Tests for a ticket
12x1,570
18,611
Where a keyword is defined
2073x167
346,452
Who calls that keyword
102x3,411
346,452
Tests with a tag
3.3x41,107
136,226
Where an endpoint is used
918x2,879
2,642,809
Method is in the README.
Two views. One engine.
Same index, for agents and for people.
Command line
| Command | Returns |
|---|---|
ticket KEY | tests that cover it |
kw NAME | where it is defined |
callers NAME | every call site |
tag TAG | tests with the tag |
find TEXT | endpoint, variable, value |
Install and run
pip install robotframework-mind[gui] python -m robot_mind_app
Windows, Mac and Linux builds: releases (unsigned, no budget for signing). On macOS: xattr -d com.apple.quarantine on the app, or build from source.

role:shopper across three repos, with kind, name, file and line.
/api/orders as a graph of 31 nodes, coloured by repo.