BlueHive HUM
Use case

Trigger AI agents via API — get the result back automatically

One HTTP POST starts an AI agent working in the background with your payload as context. HUM answers with run IDs immediately, and when the agent finishes, the signed result lands back in your system — by webhook callback or polling.

AI that fits into pipelines, not chat windows

Most "AI integrations" mean pasting into a chat box or building your own agent runtime. Your systems need a request/response contract.

Building an agent runtime is a project

Queues, retries, tool calling, timeouts, observability — running LLM agents reliably in production is months of infrastructure before the first useful result.

Fire-and-forget automation is a black box

Triggering an AI job is easy; knowing whether it worked is not. Without run IDs, status, and structured results, automations fail silently.

Glue code multiplies

Every new event source — CRM, ticketing, e-commerce, cron — needs its own custom integration to get work to an agent and results back where they belong.

How it works

How the API trigger loop works

A complete request/response cycle for background AI work.

  1. Step 1

    POST your payload

    Send JSON to your tokenized trigger URL — from a webhook, a script, Zapier, n8n, or any system that can make an HTTP request.

  2. Step 2

    Get run IDs instantly

    HUM responds 202 with a run ID and status URL for every agent the event started, so your system can correlate the work.

  3. Step 3

    The agent does the work

    Your agent runs in the background with the payload as context — searching your knowledge base and calling the HTTP tools you enabled.

  4. Step 4

    The result comes back signed

    When the run finishes, HUM POSTs the HMAC-signed result to your callback URL — or your script polls the status URL until it is done.

What you get

A production agent runtime, exposed as one endpoint

Everything you would have to build yourself — queues, signing, idempotency, observability — is already there.

Webhook-triggered agents

Tokenized trigger URLs fire custom events into HUM. Point GitHub, Redmine, your CRM, or any webhook source straight at them.

Signed result callbacks

Results POST back to your system with an HMAC-SHA256 signature, run ID headers, and per-request callback overrides.

Polling for scripts and cron

No inbound connectivity? Poll the status URL with the same token until the run leaves the queue — full output included.

Idempotency built in

Redelivered events with the same ID return the original run’s result instead of running the agent twice.

Tools and knowledge included

Agents use your org’s knowledge base and registered HTTP tools — so they can look things up and take action, not just generate text.

Run observability

Every run records status, output, tool calls, token usage, latency, and callback delivery — in the dashboard and via the agent.run.completed webhook.

FAQ

Common questions

How do I trigger an AI agent from my own system?
Create an inbound trigger in HUM settings — you get a tokenized HTTPS URL. POST any JSON to it and every event agent subscribed to that trigger’s event starts a run with your payload as context. The 202 response lists each run’s ID and status URL.
How do I get the agent’s result back?
Two ways: set a callback URL (on the trigger, or per request with the _callback field) and HUM POSTs the signed result when the run finishes; or poll the status URL from the 202 response until the status leaves "queued". Both return the agent’s output, tool-call summary, and usage.
How are callbacks secured?
Every callback is signed with a dedicated per-trigger secret: the X-HUM-Signature header carries an HMAC-SHA256 of the raw request body. Callback URLs must be public HTTPS endpoints, and both the trigger token and callback secret are rotatable at any time.
Does it work with Zapier, n8n, or Make?
Yes — use an HTTP action to POST to the trigger URL and a webhook trigger to catch the result callback. The run ID ties the two ends together, so a single Zap or workflow can send work to an agent and route the result.
What can the agent actually do during a run?
Event agents follow your written instructions and can search your organization’s knowledge base and call the HTTP tools you’ve registered — query your CRM, create tickets, send SMS, post to Slack. Each run records every tool call for review.
What happens if the same event is delivered twice?
Include an id field in your payload and HUM deduplicates: a redelivery within an hour returns the original run’s result (marked deduped) instead of running the agent again — safe for at-least-once webhook sources.
Related workflows
After-hours answeringPatient recallAppointment remindersAll featuresIndustriesPricing

Ship your first agent-powered automation today

Create a trigger, POST a payload, and watch the result come back — free for 14 days, no credit card required.