// interfaces
Dependency Updates
Start dependency update runs on demand and poll their status through the REST API without enabling a schedule.
Use the Dependency Updates API to start a run for a connected repository and poll its status. An on-demand run uses the same update generation, package risk policy, and publication safeguards as scheduled Secure Dependency Updates.
All requests require an organization API key sent as a bearer token in the
Authorization header, as described in the REST API.
Repositories and runs are scoped to the API key's organization.
Endpoints
| Method | Path | Description |
|---|---|---|
POST |
/repositories/{repositoryId}/dependency-updates |
Start an asynchronous dependency update run |
GET |
/dependency-updates/{runId} |
Retrieve an organization-scoped run and its status |
repositoryId is the positive integer ID of a connected GitHub repository. runId is
the UUID returned when a run is accepted.
Start a run
POST /api/v1/repositories/{repositoryId}/dependency-updates starts one run.
The repository must belong to the API key's organization and have an active
Superagent Security GitHub App installation. A live private run costs 5 credits
when it starts, even if it fails or finds no updates. Public repositories and
dryRun: true cost no credits.
You can start a run while the repository schedule is Off. Starting a run does not enable or change the schedule.
Request
The optional JSON body accepts scope and dryRun:
| Field | Type | Description |
|---|---|---|
scope |
string | Optional: vulnerabilities, versions, or all. Defaults to the repository's saved dependency update scope |
dryRun |
boolean | Set true to resolve updates without writing GitHub branches or pull requests. Defaults to false |
vulnerabilitiestargets known vulnerable dependencies.versionschecks direct dependencies for available versions.allchecks vulnerability remediation and regular version updates in one run.
Use {"scope":"versions","dryRun":true} to test ordinary updates separately
from {"scope":"vulnerabilities","dryRun":true}. Neither request changes the
saved scope or publishes pull requests.
An explicit scope applies to this run without changing the repository's saved
scope. Omit the body, or send {}, to use the saved scope. Requests reject
unsupported JSON fields.
curl https://superagent.sh/api/v1/repositories/1095278383/dependency-updates \
-X POST \
-H "Authorization: Bearer sk_live_..." \
-H "Content-Type: application/json" \
-d '{"scope":"vulnerabilities"}'Response
202 Accepted
{
"data": {
"id": "7b6e70c5-8cd2-4f9e-95dd-4a24e3c8a019",
"object": "dependency_update_run",
"repository_id": 1095278383,
"scope": "vulnerabilities",
"mode": "manual",
"status": "queued",
"dry_run": false,
"summary": null,
"proposal_count": 0,
"skipped_count": 0,
"published_count": 0,
"error_message": null,
"started_at": null,
"completed_at": null,
"created_at": "2026-10-01T12:00:00.000Z",
"updated_at": "2026-10-01T12:00:00.000Z"
}
}Accepted means the run has been queued, not that a pull request has been
published. If a run is already queued or in_progress for the repository,
the request returns 409 conflict rather than starting another run.
An unavailable or inactive repository or GitHub App installation returns
404 not_found. A live private run that would exceed the free-plan credit
allocation returns 402 before it is queued. The worker rechecks capacity
when the run begins; intervening usage can cause an accepted run to fail.
Retrieve run status
GET /api/v1/dependency-updates/{runId} returns 200 OK with the run under
data, using the same object shape as the start response.
curl https://superagent.sh/api/v1/dependency-updates/7b6e70c5-8cd2-4f9e-95dd-4a24e3c8a019 \
-H "Authorization: Bearer sk_live_..."Run fields
| Field | Type | Description |
|---|---|---|
id |
string | Run UUID |
object |
string | Always dependency_update_run |
repository_id |
number | Connected GitHub repository ID |
scope |
string | Effective scope: vulnerabilities, versions, or all |
mode |
string | manual for on demand runs; daily, weekly, or monthly for scheduled runs |
status |
string | queued, in_progress, completed, or failed |
dry_run |
boolean | Whether publication is disabled for this run |
summary |
object or null | Versioned, validated run evidence: alert source and target count, Dependabot job counts, candidate count, published and refreshed pull requests, withheld reasons, and error codes; independent of diagnostic logs |
proposal_count |
number | Update candidates generated by Dependabot for runs with summaries; use summary to distinguish missing evidence from zero |
skipped_count |
number | Candidates withheld from publication; reasons are in summary.withheld |
published_count |
number | Pull requests opened or refreshed during a completed live run; always 0 in dry-runs |
error_message |
string or null | Run failure information, or null when no error is recorded |
started_at |
string or null | Time execution started, or null before execution |
completed_at |
string or null | Time execution reached a terminal status, or null while active |
created_at |
string | Time the run was created |
updated_at |
string | Time the run was last updated |
Check summary, not status alone. summary.complete is false when a
Dependabot job failed, jobs were truncated, or candidates could not be
collected; missing evidence is not zero updates. summary.alerts.source is
dependabot when open Dependabot alerts were read, osv when Superagent fell
back to OSV advisories, and none for version-only runs. summary.jobs counts
one Dependabot job per ecosystem and manifest directory.
summary.withheld lists why candidates were not published:
| Reason | Meaning |
|---|---|
flagged |
The supply-chain scan confirmed a finding in the new package version |
scanner-incomplete |
The scan could not reach a conclusive result |
scanner-failed |
The scan sandbox failed |
scanner-unsupported |
No supply-chain scanner exists yet for the ecosystem |
base-moved |
The default branch moved while the candidate was assessed |
commit-failed |
GitHub rejected the update branch commit |
Older runs and runs without validated evidence return summary: null.
Timestamps use ISO 8601 strings in UTC. Poll while the status is queued or
in_progress, and stop when it is completed or failed. Back off between
requests and respect API rate limits rather than polling continuously.
Unknown runs and runs belonging to another organization return 404 not_found.
Publication is asynchronous
A completed run does not guarantee that a pull request exists. A run can
finish without eligible updates, or proposals can be skipped by policy,
the pre-PR supply-chain scan, or publication safeguards. Publication happens asynchronously after update
generation and evaluation.
To trigger automation only when an approved update reaches GitHub, subscribe
to the existing dependency_update.published organization webhook. It includes
the repository, persisted dependency update pull request ID, title, and
pull request URL. The run ID is null if no matching run was found. It is a
publication event, not a run-completion event. See
Webhooks for the payload, signatures, and retries.
MCP tools
The same operations are available through the Superagent MCP server:
| REST operation | MCP tool | Arguments |
|---|---|---|
| Start a run | trigger_dependency_updates |
repository_id (number), optional scope (vulnerabilities, versions, or all) |
| Retrieve run status | get_dependency_update |
run_id (UUID string) |
Omitting MCP scope uses the repository's saved scope. MCP has the same
organization scoping, active-run conflicts, and asynchronous publication
behavior as REST.
Errors
| HTTP status | Code | Meaning |
|---|---|---|
400 |
invalid_request |
Invalid repository ID, run UUID, JSON body, or scope |
401 |
unauthorized |
Missing or invalid organization API key |
404 |
not_found |
Repository or run is missing or belongs to another organization, or the repository or GitHub App installation is inactive |
409 |
conflict |
A repository run is already active |
500 |
internal_error |
Unexpected server failure |