Introducing Agents.·Read
OverviewAgentsWorkflowsExecutionDeveloper ToolsSecurityLibrary

Agent

A goal in full stride.

Give it a goal. It plans, composes containerized tool nodes into a graph, runs them on managed infrastructure, and queries the results.

Start a new agentGet a demo

One agent. One visible graph.

Give it a goal. It composes an editable workflow graph, runs it on managed infrastructure, and leaves queryable results on the workbench.

Scan every host for CVE-2025-53770 and triage the exposed ones.
Reading targets0.2s
Applying workflow changes: 5 nodes, 4 connections0.6s
$ trickest execute --watch18s
Presenting report: ToolShell exposure1.2s
Message the agent…
AgentFable 5
illustrative · sample conversation and findings

Memory you can see.

Mid-conversation, the agent searches vault memory, reads named sections, and writes learnings back to the workflow you are on.

Resume last week's staging discovery. What scope and rate limits did we settle on?
Searching memory for "staging discovery scope"live
searchstaging discovery scope3 sections
Reading "staging-discovery-scope"
getstaging-discovery-scopeproject
Recording a project learning
learnTRICKEST.md
Message the agent…

illustrative · sample memory transcript

  1. Search in the thread

    Semantic recall shows up as Memory tool calls in the chat timeline, next to every other action.

  2. Learn on the workflow

    Each graph keeps a project file: purpose, learnings, known issues. The next session starts informed.

  3. Stays in your vault

    Per-vault memory persists across sessions. Scope, constraints, and context survive the next chat.

Control you keep.

Authority to act never means acting unobserved. The graph stays yours to review, edit, govern, and inspect.

Reviewable

Every workflow the agent proposes is visible on the workbench before it runs. Review the graph, then approve the run.

Editable

The graph the agent builds is the same graph you can edit. Inspect a node, swap a module, or rewind a step.

Governed

Work runs in your own vault. Per-vault memory and access stay scoped to that workspace.

Inspectable

Every step logs its inputs, outputs, and duration. Runs are versioned, so you act on what changed.

questions

Frequently asked

Still deciding if it fits? Get a demo.

It executes code, works with files, and builds apps in a sandbox. It also composes and runs workflows across the fleet, then queries their results.

It can install tools, run analysis, and serve an app preview from its sandbox. It also builds workflows and turns their output into queryable tables. Read the sandbox guide.

Those help you write code or answer questions. This one plans a multi-step workflow, runs real tools on managed infrastructure, and persists the results in your workspace.

Workflow runs and detached sandbox commands continue after you close the tab. Return to their logs while the work continues. Sandbox commands and previews still depend on the sandbox remaining available.

Work runs in your own vault, and memory stays scoped to that workspace. For platform-billed requests, model providers are barred from training on your data. See AI Policy.

In the web app, Slack, the SDK, and over MCP. The same agent and workspace across all of them. Or Read the SDK docs.

Explore the platform

Adjacent surfaces on the same engine. Pick the next cut of the stack.

Get a personalized demo

See Trickest in Action

A 30-minute walkthrough. We map the platform to your stack and answer pricing and deployment questions for your environment.

Book a demoExplore the platform