Skip to main content
No. Agents are an orchestration and triage layer — they handle the repetitive work (re-running jobs, searching candidates, grouping mismatches) and surface decisions to a human reviewer. Approvals, fixes, and rejections stay with you.
Not in v1. The agent proposes match candidates as pending items; a human approves. Confidence-threshold auto-approval is on the v2 roadmap. See Rules and decisions.
Every mutating action is reversible from the workspace Undo Center for 7 days. Mismatch report state changes, mode preference changes, and pending item creations are all undoable. See Audit and undo.
Your inventory, mapping results, and pending items are sent to the configured LLM provider (Anthropic by default) only as context for the cycle. We do not train on your data. If you bring your own API key, the data goes directly to your account; we never see it.
Hosted Anthropic usage is included in the Pro Max subscription up to your configured daily spend cap (default $5/agent/day). BYO key paths bill against your own account. There’s no per-action fee.
Yes — instantly, from the agent header, via the API, or via MCP. The current cycle finishes; no new cycles start until you resume. Status PAUSED_USER.
Pro Max and Enterprise only. Basic and Professional show an upgrade card when you visit /agents. If you downgrade off Pro Max, your agents move to PAUSED_USER with config intact — re-upgrading reactivates them.
Active agents move to PAUSED_USER with reason “Plan downgraded — Pro Max required.” All config (rules, channels, data sources, memory, run history) is preserved. Re-upgrading restores everything as-is.
Yes — per-agent hourly action cap (default 60 mutating tool calls per hour, hard pause). Read tools are not counted. The cap is enforced server-side via a Redis sliding window. See Cost and limits.
Not in v1. Triggers are limited to the built-in set: inventory upload completed, mapping job finished, mismatch reported, golden-dataset changed, cron tick, data-source sync completed, user message. A customer-defined trigger DSL is on the roadmap.
No. Agents run in the same Spring API process that serves REST traffic, on the existing single-region deploy. Each agent’s cycles are serialised via a Postgres advisory lock keyed by agent_id; cross-agent parallelism is bounded by agent.runtime.max-concurrent-cycles (config, default 8). Multi-region active-active is not on the v1 roadmap.
Not directly in v1. The system prompt is composed by the runtime from agent.name, agent.goal, agent.scope_json, the current rule set, and the tool registry. You shape behaviour through the goal field and the rule set. Direct prompt editing is on the v2 roadmap.

​
Still stuck?

Email support@mapping.travel — include your workspace id and (if applicable) the run id or pending item id.