GitHub Repository Search

Build a repository shortlist for developer-tool research or open-source inventory without collecting personal contact data.

Example

{
  "query": "topic:data-visualization",
  "sort": "stars",
  "maxResults": 20,
  "maxRequests": 3
}

Results go to the dataset. Key-value records: OUTPUT explains completion; RESULTS_CSV contains the rows from this run; CHECKPOINT can be supplied on the next run with the same source inputs. A completed checkpoint performs no requests. CSV string cells that could be interpreted as spreadsheet formulas are prefixed with an apostrophe.

Completion and limits

Public repository metadata only. No private repositories, code download, personal emails or release monitoring. GitHub search exposes at most 1,000 matches; narrow the query. Shared unauthenticated quotas can stop a run.

All runs stop at the output/request limits, a 4 MiB response limit, a source error, or an invalid checkpoint. A valid zero-row response is distinct from a failed source. There are no automatic retries, proxy rotations or redirect fallbacks. A checkpoint remembers delivered IDs and the source cursor; delivery is acknowledged before an ID is recorded. A process crash between dataset delivery and checkpoint persistence can still repeat rows. Mutable upstream results can move during pagination. This is best-effort continuation, not an immutable snapshot, exactly-once delivery, or a change-monitoring service. Reset the checkpoint for a fresh export. Checkpoints are capped at 10,000 source IDs.

Attribution and data rights

Repository metadata includes the declared SPDX identifier when present. This is not a legal license audit; repository source code is not downloaded.

Official source guidance: https://docs.github.com/en/rest/search/search#search-repositories

What can this tool do?

Build a repository shortlist for developer-tool research or open-source inventory without collecting personal contact data.

What data does it return?

Source-qualified records: id, repository, url, stars, forks, language, license, description. Exact missing values remain null or absent as described above. JSON/CSV can include nested fields as JSON cells.

How to use it

Use example-input.json. Start with a small output limit. Read OUTPUT.complete and OUTPUT.reason before treating the export as complete. Continue with CHECKPOINT only when needed; omit it for a fresh snapshot.

Pricing

Actor fees: $0.00005 per run start, plus $0.0002 per delivered repository ($0.20 per 1,000 results). No separate automatic dataset-item fee is intended.

For 100 delivered repositories, the Actor fee is $0.02005, including one start. Source errors and genuine empty results can still incur the run-start fee. These are Actor fee examples, not a promise of total platform cost or creator earnings. Source APIs are free within their rules; operating this tool through Apify can consume platform credits. Check the Store price and your account limits before running.

Set a small result cap and an Apify run spending limit. The exporter checks available result-event allowance before requesting and delivering data; it stops with a saved summary and checkpoint when the allowance is exhausted. A strict zero run limit produces no source request. Interrupted runs and checkpoint recovery can still repeat a delivered row; inspect the saved output before resuming.

Is it legal?

Use the official source API within its rules. Preserve the attribution and license fields described above. Do not use this tool to circumvent authentication, restrictions or third-party rights. Source guide: https://docs.github.com/en/rest/search/search#search-repositories

FAQ

Does zero rows always mean no matching data? No. A source error is reported separately and the run fails after saving a partial summary. A genuine empty source response can complete normally.

Can it resume without duplicates? It suppresses delivered source IDs while a checkpoint remains valid. Process crashes or upstream search changes can still lead to repeats or omissions; see the checkpoint limits above. It does not promise exactly-once delivery or monitoring.

Feedback

Record the exact source input, completion summary and expected result when evaluating a problem; never share API keys or confidential source data in a public issue.