When something goes wrong in a multi-agent system, the hardest part is usually not the diagnosis — it is knowing where to start looking. Global search solves this directly: type what you know — a task ID, an agent name, a connection label — and results surface across agents, tasks, connections, workflows, and audit logs from a single entry point, without navigating to each section first.
The problem with siloed navigation
Most platforms organize information by type: a page for agents, a page for tasks, a separate view for connections. That structure works well when you know which category you want. It breaks down during investigation, when you have a partial identifier — a task ID in an error message, an agent name from a support ticket, a connection label from a budget alert — and you do not know which section of the interface to open.
The cost of siloed navigation adds up. Each wrong section is a page load, a search, and a confirmation that the result is not there. Multiply that by the urgency of an incident and the number of team members investigating simultaneously, and the friction becomes significant. A global search that spans every resource type collapses that search space into a single query.
What global search covers
A well-designed global search surface returns results from every first-class resource type in the platform: agents, tasks, workflow runs, connections, MCP servers, and audit log entries. Results group by resource type so you can see at a glance whether your query matched an agent name, a task reference, or a log entry — without the types bleeding into each other.
The scope of search matters as much as its speed. A search that covers agents but not their associated tasks is only marginally better than navigating directly. When the search spans the full estate, you can start with any piece of information you have — a partial name, a reference ID, a date — and arrive at the right resource regardless of what type it turns out to be.
Praesidia scopes every search result to your organization automatically. Results from other tenants are never included regardless of query content, which matters both for data isolation and for keeping results relevant: the set you are searching is exactly your fleet, not a shared global index.
Keyboard-first access
Global search is designed to be reached without lifting your hands from the keyboard. A keyboard shortcut opens the search overlay from any screen in the platform. The same overlay accepts a typed query and returns live results as you type, so the fastest path from "something is wrong" to "I found the relevant record" is a key press and a few characters.
This matters in practice. Operators investigating a live incident are already cognitively loaded. Reducing the interaction to a keyboard shortcut and a typed query keeps them in flow. The alternative — clicking through navigation, opening the right section, then searching within it — introduces friction at exactly the moment you need the least of it.
Result quality and ranking
Raw full-text search over a large estate produces too many results to act on. Useful search needs ranking: results that match exactly on a unique identifier should appear above results that match a common word somewhere in a description. Results from recently active resources are generally more relevant than archived or inactive ones.
Name and identifier matches rank above description and metadata matches. Within a type, recent activity is a relevance signal. The result set is bounded so that the overlay remains responsive even as the estate grows — you see the most relevant matches immediately rather than waiting for an exhaustive scan.
Exact-match queries on structured identifiers — task IDs, agent slugs, connection names — return precise results quickly. Freetext queries against names and descriptions return broader results ordered by relevance. Both patterns are useful in different investigation contexts.
Search within sections
Global search is complemented by within-section filtering on every list view in the platform. The agent list has its own search. The task history view has its own filters. These are not redundant — they serve a different purpose. When you already know you are looking at agents and want to filter by name or status, in-section search is faster because results are already constrained to the right type and the filter controls are already visible.
The two surfaces work together: global search gets you to the right section, and in-section filtering narrows within it. An investigation flow often uses both: global search to locate an agent by a partial name, then the agent's task history filtered by date to find the specific run you are examining. Neither surface replaces the other; they cover different points in the same workflow, and together they make the full estate navigable without requiring you to know the layout in advance.
Access control and search scope
Search results respect the same permission model as direct navigation. A user who cannot see a resource by navigating to it also cannot find it through search. Access control is not a separate layer applied on top of search results — the same organization-scoped access model that governs every part of the platform governs search results too.
This matters for multi-role teams. A billing analyst who has permission to view cost data but not agent configuration details will see cost-related results and not configuration records. The search surface reflects their actual permission set rather than a separate visibility model that could diverge over time. When permissions change — because a user changes roles or a team membership is updated — search results reflect the updated access immediately, so the search surface stays consistent with the rest of the platform.
Global search is one component of a broader observability picture. For the full view of what is happening across your AI estate, see Observability for AI Agents: Logs, Metrics, and Traces and The Operations Dashboard for Your AI Estate. For teams managing many agents across a growing estate, building an AI agent inventory is the complementary discipline — knowing what you have is a prerequisite for finding it quickly.
Common questions
Does global search include audit log entries?
Yes. Audit log entries are part of the searchable estate, subject to your permission to view audit data. This means you can locate a specific log entry by actor, action, or resource reference without navigating to the audit log section and rebuilding your filter from scratch.
Is search scoped to my organization automatically?
Yes. Every query is scoped to your organization's data. You will never see results from other tenants, and there is no option to broaden the scope beyond your own estate. This scoping is a structural guarantee, not a filter applied after results have been assembled.
How do I access global search?
Global search is accessible from any screen via a keyboard shortcut, or by clicking the search control in the top navigation bar. Results appear as you type. Selecting a result navigates directly to the corresponding resource. For a broader look at how search fits into day-to-day operations, see Saved Views for Faster Operations.