the execution layer

AI reasons. Trickest executes.

A hyperscalable engine, pre-integrated tools, and queryable tables behind one API and one bill. No platform team to hire.

Run your first workflow

Get a demo

The agent gap

Agents plan. They fail at deterministic runs.

Reasoning picks the next step. Enterprise still needs a system that runs the same workflow safely, visibly, and at scale.

The engine that makes the plan run.

Agents pick the next step. Trickest builds the graph, bursts the fleet, and leaves typed output you can query. Same canvas humans edit by hand.

The agent builds the graph.

Same canvas you edit by hand.

Burst a node across the fleet.

Machines appear, then vanish.

Live tables from any output.

Typed rows from tool output.

Workflows become apps.

Dashboards from live output.

Every agent gets its own sandbox.

One sandbox per agent. Tools, files, and runs stay isolated.

What teams shipped in production

Anonymized outcomes from customers running Trickest at scale.

Financial Technology
Adversary simulation dropped from three weeks to six hours per tactic.
Fast-Moving Consumer Goods
Security testing scaled from eight to thirty tests a month with the same team.

Read the case studies

What security teams run

Flagship programs built on the same engine: continuous, scoped, and reproducible.

Connect

Start where you sit.

trickest · connect

$ curl -fsSL https://trickest.io/install.sh | sh
$ trickest auth login
$ trickest attach
$

Create an account CLI docs

@trickest/sdk

The typed client your agents already know how to call.

One npm package, native fetch, no HTTP client to wire up. The whole platform behind a typed API that your coding agent can reach through the CLI, SDK, and MCP.

src

lib

src›lib›runs.ts

1import { TrickestClient } from "@trickest/sdk"
2 
3const t = new TrickestClient({
4  token: process.env.TRICKEST_TOKEN,
5})
6 
7// Boot a full workflow run on a managed fleet.
8const run = await t.runs.execute(workflowId, {
9  fleet: "asm-fleet",
10  inputs: { domain: "northstar.io" },
11})
12 
13// Re-execute one node. Upstream deps memoized.
14const partial = await t.runs.execute(workflowId, {
15  node: "httpx",
16})
17 
18// Pull a single node's output files.
19const output = await t.runs.getNodeOutputFiles(
20  run.id,
21  "findings",
22)

app · findings

Explore additional resources

Field notes and docs for teams building on Trickest.

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 demo Explore the platform

import { TrickestClient } from "@trickest/sdk"

const t = new TrickestClient({

token: process.env.TRICKEST_TOKEN,

})

// Boot a full workflow run on a managed fleet.

const run = await t.runs.execute(workflowId, {

fleet: "asm-fleet",

inputs: { domain: "northstar.io" },

})

// Re-execute one node. Upstream deps memoized.

const partial = await t.runs.execute(workflowId, {

node: "httpx",

})

// Pull a single node's output files.

const output = await t.runs.getNodeOutputFiles(

run.id,

"findings",

)