Agent dashboard icons

Icons for watching agents work — runs, steps, tool calls, hand-offs, approvals, budgets and the moments they need a human.

An agent dashboard shows things no other software has needed an icon for until now: a run that is thinking, a tool being called, a hand-off between two agents, a step limit, a human approval gate, a spend cap. Drawing these one at a time, screen by screen, is how a product ends up with three different agents on one page.

This collection is the agent vocabulary as one family: the same figure, the same octagon, the same marks. Pair it with the ordinary controls — play, pause, stop, retry, log — which are drawn on the same grid and in the same weights.

import { AgentRun, ToolCalling, HumanApproval } from "@iconmind/react";

The 39 icons, described

Agentagent: An autonomous AI agent that decides and acts on its own to reach a goal, without step-by-step prompts.
Agent runagent-run: Run an agent — start an AI agent on its task and let it work through its loop.
Agent pauseagent-pause: An agent with a pause mark beside it — the agent is held where it is, suspended but not stopped.
Agent stopagent-stop: Stop an agent — halt a running AI agent before it takes another step.
Agent thinkingagent-thinking: An agent thinking — the AI is busy reasoning, planning or working on the task.
Agent workingagent-working: Agent working — the agent is busy on the task right now and will report back.
Agent idleagent-idle: Agent idle — ready and waiting, with nothing to do right now.
Agent doneagent-done: Agent done — the agent finished its task, and it went well from start to end.
Agent alertagent-alert: An agent with an alert mark beside it — something is wrong with this agent and needs attention.
Agent blockedagent-blocked: Agent blocked — the agent cannot go on, stuck waiting on something it does not have.
Agent handoffagent-handoff: An agent handoff — one agent passing its work and context on to another.
Agent delegateagent-delegate: Agent delegate — hand a piece of work down to a smaller sub-agent.
Agent teamagent-team: An agent team — a standing group of agents that work together as a crew.
Agent swarmagent-swarm: An agent swarm — many agents working loosely together as a distributed crowd.
Supervisorsupervisor: A supervisor agent — the orchestrator that directs other agents and decides who does what next.
Subagentsubagent: A subagent — an agent spawned to work for another agent on a piece of the task.
Tool callingtool-calling: Tool calling — an agent reaching outside itself to run a function, an API or a tool.
Tool resulttool-result: A tool result — what the tool handed back to the agent after it was called.
Tool errortool-error: A tool error — the tool broke instead of answering, an exception.
Human approvalhuman-approval: Human approval — a gate where a person has to say yes before an agent continues.
Escalate to humanescalate-human: Escalate to human — past what the agent should decide alone, handed up to a person.
Step limitstep-limit: A step limit — only this many tries or iterations before an agent must stop.
Turn limitturn-limit: A turn limit — the conversation has a budget of turns and then it stops.
Spend capspend-cap: A spend cap — the money stops here, a hard ceiling on the budget.
Agent costagent-cost: An agent beside a coin — what a single run of this agent costs to operate.
Agent logagent-log: An agent log — everything the agent did, recorded in order as a trace you can read back.
Tracetrace: A trace — one request followed through every service it touched, span by span.
Spanspan-trace: A span — one timed piece of work inside a distributed trace across services.
Checkpointcheckpoint: A checkpoint — a model's weights saved exactly as they were at this moment in training.
Replanreplan: Replan — that plan is no good, discard it and make another before acting.
Retryretry: Retry — go round and try the same operation again after a failure.
Loop detectionloop-detect: Loop detection — the agent is going round in circles and has been caught.
Episodic memorymemory-episodic: Episodic memory — what happened, in order, an agent's record of events.
Memory pinmemory-pin: Memory pin — this memory is not allowed to fade, pinned to be kept.
Goalgoal: A goal — the target or objective an agent, a team or a project is working toward.
Objectiveobjective: An objective — the target a model, an agent or a team is trying to reach.
Agent task queuetask-queue-agent: An agent task queue — what the agent will do next, in order of arrival.
Pausepause: Pause — hold it there, suspend playback or a running process without ending it.
Stopstop: Stop — end it now, halt playback or cancel a running process.

In code, each is one import — import { Agent } from "@iconmind/react/icons/agent" — and the same name in Vue, Svelte, Solid, Preact, React Native, Astro, Blade and Flutter.

Other collections

AI chat interface RAG pipeline MCP servers and tools LLM observability Voice agent CI/CD pipeline Data platform Security posture Cloud console Automation builder Analytics dashboard Devices and hardware

Every agent dashboard icon, free to ship

39 icons drawn on one grid, in outline and duotone at three weights, with the code for nine frameworks. MIT — commercial use, no attribution, no seat count.

Browse the whole set Install the package Request an icon