AXME
Orchestrate work across services, agents, and humans — without webhook glue
Coordinating work across 3+ services with approval steps requires: REST calls, webhook endpoints, polling loops, state reconciliation, and retry logic. Every integration is custom.
AXME Cloud's intent protocol treats every participant (API, agent, human) as an equal peer.
Coordinating work across 3+ services with approval steps requires REST calls, webhook endpoints, polling loops, state reconciliation, and retry logic. Every integration is custom.
Microservices plus agents plus humans
A customer onboarding flow touches CRM, billing, identity, and an LLM agent — with legal approval in the middle. Each service team owns REST APIs and webhooks; platform owns the nightmare of cross-service state.
AXP intents let every service and agent participate as peers on one durable lifecycle — humans included — without a bespoke state machine per product line.
Example: B2B onboarding
Agent verifies documents (service A), creates billing account (service B), waits for legal approval (human), provisions tenant (service C). Today: choreographed sagas with compensating transactions scattered across repos.
With AXME: one intent coordinates waits; each service completes its leg via SDK; partial failure retains audit and retry points.
Webhook glue → intents
Bespoke integration
# service A webhook → service B poll → manual approval
AXP coordination
intent = axme.submit(flow) await intent.wait_for_human(...) await intent.complete()
SOLUTION
How teams solve this with AXME.
APIs as peers
Services submit intents.
Humans in the flow
Approval mid-chain.
Partial completion
Track per-step state.
Service integration pattern
Expose small AXME adapters in each service — submit child operations or complete tool waits. Keep domain logic in services; keep coordination and HITL in intents.
Use multi-tenancy namespaces per customer for isolation in SaaS onboarding flows.
Common questions
- Replace Temporal sagas?
- Some teams replace; others run AXME for agent+HITL legs and keep Temporal for deterministic backend sagas — evaluate per boundary.
- Synchronous REST still?
- Yes for request/response; intents cover async and human-delayed legs.
- Versioning intents?
- Version intent schemas in your API contract like any cross-service interface.
Related reading
Deeper dives from the AXME blog.
I Stopped Building Webhook Retry Logic. Here's What I Use Instead.
Exponential backoff, jitter, dead letter queues, idempotency keys, HMAC verification - all to deliver one message reliably. There are better options now.
Read post →A2A Tells Agents How to Talk. It Doesn't Tell Them What Happens When Things Break.
Google's A2A protocol handles agent communication. But crash recovery, retries, timeouts, and human approval gates? That's still on you. Unless you add a lifecycle layer.
Read post →
Related capabilities
Related links
Ship your first durable agent — in under 10 minutes.
Free tier. No credit card. Self-host or hosted — your choice.