API Reference

Add a single item to a dataset's draft by hand.

Remove duplicate draft items identified by a preview.

Compare two published versions and see what items were added, removed, or changed.

Create a dataset.

Fetch a single dataset by its UUID.

Fetch a single draft item, including its input/output payload and revision history.

Add an item to the draft from a completed workflow run, capturing its actual input and output.

Bulk-import items from a CSV file, mapping columns to input/expected-output fields.

List a project's datasets.

List the items currently in a dataset's draft.

List a dataset's published versions.

Preview which draft items would be removed by deduplication, without changing anything.

Publish the current draft as a new immutable version.

Reset the mutable draft back to the contents of a published version.

List the functions, workflows, and agents registered by a deployment's workers.

Read application logs emitted by a deployment's workers.

List workflow runs executed on a deployment.

List the worker processes currently executing a deployment's workflows.

Fetch a single deployment by its UUID.

List deployments the authenticated user can access.

Create a custom environment beyond the four defaults, for example a per-feature preview environment.

Fetch a single environment by its UUID.

List a project's environments.

Cancel an in-progress run.

Cancel an experiment and every run currently in progress under it.

Create an experiment against a dataset.

Fetch a single run by its UUID.

Fetch a single item's actual output, expected output, and per-scorer results.

Fetch a single experiment by its UUID.

Fetch a run's aggregate results once it completes.

Fetch the full evidence (prompt, model reasoning, or code output) behind a single score.

Fetch links back to the durable execution journal entries behind a run's items, for step-by-step debugging.

List a project's experiments.

List the runs triggered for an experiment.

List individual scorer results across runs, for building custom dashboards.

Fetch the pass/fail gate result for a run, for use in CI pipelines.

List the dataset items a run scored below threshold on.

List the scorers applied to a run and their aggregate results.

Fetch a run's pass/fail counts, average scores, and duration.

Trigger a new run of an experiment against its current dataset version.

Fetch a single project by its UUID.

List the projects visible to the authenticated user, across every workspace they belong to.

Open a quality case, optionally linked to the run or score that surfaced it.

Fetch a single quality case by its UUID.

List a project's quality cases.

Fetch a single run summary by its run ID.

Fetch every span recorded for a single trace.

List run summaries: status, timing, retries, and LLM cost for each run.

List distributed traces recorded for a workspace, optionally scoped to a single deployment.

Read structured logs emitted during a run, across every worker and step.

Fetch a single scorer by its UUID.

List a project's scorers.

Create a secret. The plaintext value is encrypted at rest and never returned by any endpoint — only its masked form.

Delete a secret. Deployments that reference it lose access to the value on their next restart.

Fetch a single secret's metadata by its UUID.

List secrets visible to the authenticated user. Values are always masked; use the reveal endpoint to read a decrypted value.

Update a secret's name, type, or value.

Fetch a single worker by its UUID.

List workers. Scope the query with either project_id or deployment_id.

Authenticate requests to the AGNT5 API with a personal API key or a bearer token, and understand the error envelope and rate limits every endpoint shares.

List the projects your credential can access and fetch a single project's metadata, workflow count, and production deployment.

List, fetch, and create the environments a project deploys into: local, preview, staging, and production.

Store, list, and rotate the credentials your deployed workflows read at runtime, scoped to a project and environment.

List and inspect deployments, and read their logs, workers, components, and runs.

List and inspect the worker processes that execute your deployed workflows, agents, and functions.

List workflow runs, read a run's logs, and fetch the distributed trace recorded for each run.

Create eval datasets, publish immutable versions, import examples from runs or CSV, and deduplicate items.

Create experiments, trigger runs against a dataset, and read run results, failures, and scores.

List and fetch the scorers a project uses to grade experiment results.

List, fetch, and open quality cases: tracked behavior regressions linked back to the runs, scores, and datasets that surfaced them.