OpenGuilds is the human-in-the-loop layer for modern AI. Send tasks via REST, and verified guild workers — backed by AI fallback — guarantee every task completes. No null results. No stalled pipelines. Ever.
POST /v1/tasks
task_type: "classification"
engineeringlabel: "bug"
confidence: 0.91
passed200 OK
latency: 8.4s
deliveredSend a single JSON payload to /v1/tasks. Pick a guild, a task type (classification, annotation, judgment, RLHF), and a confidence threshold.
Verified guild workers complete the task in parallel. Each answer carries a confidence score; we compute a weighted result and route low-confidence tasks for review.
Once consensus is reached — or AI fallback completes — we POST the result to your webhook with a signed payload and a completion_source field.
Every task is guaranteed to complete. Verified humans answer first; AI agents step in if a task crosses your timeout. You never get a null result.
Multiple workers complete each task in parallel. We return a result only when their agreement passes the confidence threshold you set.
POST a task, get a webhook. No queues, no orchestration, no SDK gymnastics. Python and Node SDKs are optional.
Two guilds open today. Four more in development.
Content moderation, data labeling, image classification, form extraction, opinion surveys.
Code review, security triage, technical documentation QA, architectural judgment.
Clinical note review, ICD-10 coding, symptom classification, medical image annotation.
Contract review, case law analysis, compliance evaluation, legal document classification.
Climate data labeling, carbon validation, environmental impact classification, policy QA.
Translation QA, cross-lingual moderation, multilingual RLHF preference ranking.
Free tier includes 100 tasks per month. No credit card. Set up your first task in five minutes.