Does it write anything to my GitHub account?

No. Tracer only reads things that are already public. It never comments, never assigns, never forks, and never opens a pull request as you. The permission it asks GitHub for covers your public profile and email address, and nothing else.

Does it need AI to work?

No. The scoring is worked out from what is actually on GitHub, and the same issue always gets the same score. There is an optional AI reading you can ask for on any issue, and everything else works exactly the same with it turned off.

Nothing. It is free, and the code is open, so you can also run your own private copy if you would rather.

How is the score calculated?

Eight things are measured and weighted, and the whole method is written down in the open. If you think a weight is wrong, it is one file, and you can suggest a change to it.

Never contributed before? Start here

Find open-source issues actually worth your time

Searching good first issue gives you a million results and no way to choose. Tracer scores them against what you know and the time you have, then tells you which ones to take.

Read the source

Free, open source, and read-only. Nothing is written to your account.

Your queue · 6 scored against your profile

All6Take it2Worth a look1Skip3

79%
Make the right sidebar resizable, like the left one
openstreetmap/iD
Your stack, nobody assigned, maintainers active this week
JavaScript3-10 hours
71%
Improve error message when config is missing
acme/runtime
Small and self-contained, and the issue says what finished looks like
TypeScript1-3 hours
58%
Add tests for the pagination helper
acme/utils
Good practice, but outside pull requests have not merged in 90 days
TypeScript2-5 hours
31%
Add a dark mode toggle to the settings panel
acme/design-system
Someone called it in the comments eight months ago
24%
Update the installation guide for the new CLI
acme/cli
A linked pull request already merged. The issue was never closed
19%
Fix toolbar icon alignment on Safari
acme/toolbar
Open since 2017, no comment since 2019

Two worth taking, one worth a look, three to skip. Each with the reason.

Eight signals per issueEvery score explainedNo AI requiredMIT licensedRead-only access

A label tells you almost nothing

Maintainers add labels once and rarely revisit them. Everything that decides whether an issue is worth taking happens after that, in the thread and in the dates.

It reads the comments, not the label

Half the issues that look free are not. Somebody says they will take it in comment seven, nobody assigns them, and the label never changes. Tracer reads the thread and tells you before you spend an evening on it.

Comment 7 of 31

@contributor

I'd like to work on this one, assigning myself

8 months ago. Never assigned.

Someone called it

Skip it

It knows whether work actually lands

A project can look busy and still merge nothing from outsiders. Tracer measures real throughput: how many outside pull requests merged, how long a first review takes, how much is rotting in the queue.

Merged from outsiders, 90 days: 28
First reply to a pull request: 14h
Open longer than 90 days: 2
Last commit: today

Work lands here

It shows the whole calculation

Eight weighted signals, each with the reasoning that produced it, and the arithmetic on screen. Nothing hides behind the number, and the weights live in a file you can open a pull request against.

Your stack81 × 25% = 20.3

Actually free95 × 20% = 19.0

Project health72 × 15% = 10.8

Issue clarity44 × 15% = 6.6

+ four more signals = 79%

Two minutes to set up

Then the queue keeps itself current, and re-checks anything you save in case somebody else gets there first.

1
Sign in with GitHub
Read-only. Tracer never writes to your account, and asks for nothing beyond your public profile.
2
Say what you know
Languages, frameworks, what you want to learn, and how much time you actually have this week.
3
Get a ranked queue
Issues scored against your profile, each with a verdict and the reasons behind it. Save the good ones, hide the rest.

You can check the working

A recommender nobody can take apart is not worth trusting. Here is the entire model, live from the code that runs it. Disagree with a weight and you can open a pull request against it.

Read the scoring model

Your stack: 25%
Actually free: 20%
Project health: 15%
Issue clarity: 15%
Your time: 10%
Learning: 5%
Maintainers: 5%
Competition: 5%

Questions people ask

Does it write anything to my GitHub account?: No. Tracer only reads things that are already public. It never comments, never assigns, never forks, and never opens a pull request as you. The permission it asks GitHub for covers your public profile and email address, and nothing else.
Does it need AI to work?: No. The scoring is worked out from what is actually on GitHub, and the same issue always gets the same score. There is an optional AI reading you can ask for on any issue, and everything else works exactly the same with it turned off.
What does it cost?: Nothing. It is free, and the code is open, so you can also run your own private copy if you would rather.
How is the score calculated?: Eight things are measured and weighted, and the whole method is written down in the open. If you think a weight is wrong, it is one file, and you can suggest a change to it.

Find something worth doing this weekend

Set up a profile in two minutes and get a queue that is actually yours.

FreeRead-onlyNo AI requiredSelf-hostable