AI-native platform
Trace AI loads tenant playbooks, reads operational data on the tenant, proposes actions, and waits for approval before anything executes. Integrations, ProcessFlow, Datapool, and WebApps share one policy boundary—so you ship AI-assisted ops without losing audit or control.
Integration recovery assistant
Policy-bound · Skill: failed-sync-retry
Simulated conversation — integrations execute only after human approval
Monitor alert → Trace AI triage → human approval → governed retry → process resumed
AI-native platform
Trace AI, LLM steps, and MCP tools run inside the same tenant as ProcessFlow, integrations, and Datapool. Every mention of AI here comes with a concrete outcome for your team—not a feature checkbox.
Trace AI reads Monitor history, granted Datapool rows, and integration runs already on your tenant. No copy-paste into a sidecar chat window.
Ops answers in minutes: failed sync → proposed retry → approval → resume process.
Before execute_tenant_integration or risky writes run, the agent pauses for human-in-the-loop. Guardrails and sandbox profiles apply to every tool call.
AI speed without rogue API calls or unaudited side effects.
Tenant SKILLS/ define how agents triage, escalate, and respond—versioned runbooks ops owns, not prompt drift in a consumer chat UI.
Consistent handling across shifts, channels, and handoffs.
Classification, OCR, extraction, and generation run as governed step code alongside PHP, JavaScript, and TypeScript—with the same Monitor trail.
Document-heavy and judgment steps automate without standing up another pipeline.
See the full agent story on AI, Agents, and MCP—including live Monitor-to-retry demos and tenant SKILLS/ playbooks.
Six pillars on one tenant—each wired so Trace AI, LLM steps, and MCP tools work on real integrations and data. Select a pillar to see what that means for your team.
Triggered from chat, WebApp, process, scheduler, or event. Loads tenant skills, reads granted data, proposes actions, and pauses for approval before execute_tenant_integration runs under guardrails.
From Monitor alert to governed retry: triage on tenant data, wait for approval, execute, resume the paused process—full trace linked to the incident.
Triggered from chat, WebApp, process, scheduler, or event. Loads tenant skills, reads granted data, proposes actions, and pauses for approval before execute_tenant_integration runs under guardrails.
Trace AI · skills
HITL · MCP tools
Developer APIs with MCP, governance for agent tool calls, and observability for model spend—so AI ships inside one policy boundary from first commit to production incident.
/api/v1/* OpenAPI · MCP
REST API, OpenAPI, scoped keys, MCP, and AI-assisted solution development.
REST API, OpenAPI at /api-docs/, scoped keys, and MCP tools so Cursor, Claude, and Lovable scaffold against real tenant APIs—agents ship with Trace AI, skills, and governed execute from day one.
Ship agent-ready solutions from Cursor or Claude—MCP and OpenAPI point at governed tenant APIs, not mocks.
Learn moresandbox · keystore scoped API keys
Guardrails, keystore, sandbox profiles, API scopes, and human approval paths.
Sandbox guardrails, keystore, scoped API access, MCP deployment rules, and human approval paths—so AI tool calls stay inside policy, not prompt hope.
Every agent tool call runs inside sandbox and MCP guardrails—policy is the control plane, not the prompt.
Learn moreMonitor · queues 9 log types
Dashboard triage, Monitor logs, queues, worker health, audit trails, and LLM ops.
Dashboard triage, Monitor with nine log types, actionable queues, worker health, DataPool audit, and LLM spend control—see every agent action, approval, and model cost in one tenant.
See model spend, agent approvals, and integration outcomes in one Monitor view—triage AI like any other production workload.
Learn moreCopilots that cannot see your processes or data
Trace AI on granted Datapool and integrations—proposes retries and updates on live tenant state, not pasted exports.
AI that acts before anyone approves
Human-in-the-loop before execute_tenant_integration—speed with guardrails, sandbox profiles, and audit on every tool call.
Prompt drift across shifts and channels
Tenant SKILLS/ playbooks ops publishes—consistent triage whether the trigger is chat, WebApp, or Monitor alert.
Document AI as a separate pipeline
OCR and LLM inside ProcessFlow steps—same Monitor trail as the rest of the workflow.
Integration sprawl without ownership
One connector model for ProcessFlow, API, and agents—with history and audit per run.
No visibility into model spend or agent actions
LLM ops in Monitor—token usage, approvals, and execution linked to incidents and processes.
Trace AI assists and proposes. ProcessFlow orchestrates with LLM and OCR steps. Integrations connect after approval. Monitor shows agent reasoning alongside every run.
01
WebApps, API, schedule, webhook, or Trace AI chat starts work on the tenant.
02
Trace AI loads tenant skills, reads granted data, and proposes the next action—classification, enrichment, or retry.
03
ProcessFlow runs governed steps—LLM, OCR, branch logic, integrations, and DataPool writes in the sandbox.
04
Human-in-the-loop where policy requires it; then integrations execute with audit per run.
05
Monitor shows the full trail—agent reasoning, approvals, process steps, and integration outcomes.
See end-to-end patterns in reference workflows and solutions by team.
Start with a Monitor-to-retry agent demo, then walk integrations, ProcessFlow, Datapool, WebApps, and LLM ops—in one governed runtime aligned to your operations patterns.