--- title: "Trickest" canonical: https://trickest.com/index.md --- # Trickest > Trickest runs security workflows across a fleet and gives agents isolated environments for commands, files, and apps. Use `@trickest/sdk`, the CLI, REST API, or hosted MCP server for platform workflows. Use `@trickest/sandbox` for standalone sandbox environments. This manual covers authentication, workflow execution, queryable results, and how to choose the right SDK. ## What Trickest is A workflow engine for offensive security work. You compose containerized tools, your own scripts, and reusable modules into a versioned DAG; the platform executes it across a fleet of machines, shards a single node across many inputs to fan out, and turns JSONL outputs into live tables you query with TQL (a filter expression language, not SQL). You drive all of it programmatically — CLI, TypeScript SDK, REST API, or the hosted MCP server. Nothing here requires a browser. Trickest also provides standalone agent environments through `@trickest/sandbox`: commands, files, persistent stop/resume, and app previews. These environments have their own lifecycle; they are distinct from workflow execution machines and the platform SDK's workflow-linked sandbox service. Trickest, Inc. runs the platform at `https://trickest.io`; this site (`https://trickest.com`) is the public documentation and catalog. ## When to use Trickest Reach for Trickest when the job is **a pipeline of security tools run at scale**, and the scale or the repeatability is the hard part: - **Recon / attack-surface sweeps over a large scope** — enumerate subdomains, resolve DNS, probe web servers, and inventory ports across many root domains, then keep the inventory current on a schedule. - **Distributed vulnerability scanning** — fan one scanner (nuclei, a template set, a CVE check) across thousands of targets on many machines instead of one box. - **DAST against a defined scope** — crawl and scan an application with in-scope/out-of-scope regexes and produce a report. - **"Run this tool I already know, but across everything"** — the tool is in the library as a container; you wire inputs and let the fleet shard it. - **Reusing a known-good workflow** — the public library has ready-made DAGs (including same-week CVE checks); open one, point it at your targets, run it. - **Wiring a scan into an agent loop** — outputs land as JSONL, get promoted to live tables, and are queried with TQL, so the next step in your loop reads structured results rather than scraped stdout. For **isolated agent execution**, use the [standalone Sandbox SDK](https://trickest.com/docs/developer-tools/sandbox-sdk) to create an environment, run commands, manage files, and preview apps. This path does not require a workflow graph. The [agent sandbox guide](https://trickest.com/docs/using-the-app/ai-agent/sandboxes) explains the environment lifecycle and isolation boundary. Do **not** reach for Trickest when: - A local command already meets the task's needs for isolation, repeatability, and scale. - You need an answer, not an execution. Trickest runs tooling; it is not a knowledge base or a chatbot. - You want a workflow run but have no execution capacity. Workflow runs need a fleet with an active machine; Community workflow execution uses your attached machine (`trickest attach`). Sandbox access is a separate requirement described in its SDK guide. - You are not authorized to test the target. Scope is your responsibility, not the platform's. **How to drive platform workflows.** Use these interfaces with a static API token created in the Trickest dashboard (see https://trickest.com/auth.md): - **Hosted MCP** — `https://api.trickest.io/mcp` (Streamable HTTP), header `Authorization: Token `. Best when you can call tools natively. - **REST** — `https://trickest.io/api`, described by `https://trickest.com/openapi.json`. Best when you are writing code. - **CLI** — `curl -fsSL https://trickest.io/install.sh | sh`, then `trickest ` with JSON output. Best when you have a shell. ## Key URLs - [llms.txt](https://trickest.com/llms.txt): The routing index: this site's sections, the library, modules, tools, and the glossary as `- [Title](url): summary` links. - [llms-full.txt](https://trickest.com/llms-full.txt): The full operating manual in one file (same payload as /for-agents.md). - [for-agents.md](https://trickest.com/for-agents.md): Auth, the workflow/DAG model, the CLI grammar, TQL, and copy-paste recipes. - [SKILL.md](https://trickest.com/SKILL.md): The installable Trickest Agent Skill — first-time integrate, then the operating guide. - [openapi.json](https://trickest.com/openapi.json): OpenAPI 3.0 description of the REST API at `https://trickest.io/api`. - [auth.md](https://trickest.com/auth.md): How an agent obtains, uses, and revokes a Trickest credential. - [pricing.md](https://trickest.com/pricing.md): The two packages and what each includes. - [Documentation](https://trickest.com/docs): Product docs: key concepts, using the app, CLI, SDK, API reference. - [Workflow library](https://trickest.com/library): Ready-made workflows you can open node by node, plus every tool and module. - [sitemap.xml](https://trickest.com/sitemap.xml): Every indexable URL on this site. ## Every machine-readable surface - [OpenAPI 3.0 spec](https://trickest.com/openapi.json): The full Trickest REST API (`https://trickest.io/api`) as a spec you can code-generate from. - [Site index (markdown)](https://trickest.com/index.md): This site as one markdown page — what Trickest is, how to call it, every agent URL. Also at /llms.md. - [Credentials walkthrough](https://trickest.com/auth.md): How an agent gets a Trickest credential, what the 401 looks like, and how it is revoked. - [Pricing](https://trickest.com/pricing.md): Both packages, what each includes, and the honest limits — in markdown, no numbers invented. - [AI catalog](https://trickest.com/.well-known/ai-catalog.json): Machine-readable capability catalog for this site. - [Agent card](https://trickest.com/.well-known/agent-card.json): Agent-facing identity and entry points for Trickest. - [Docs MCP server](https://trickest.com/api/mcp): Read-only, unauthenticated MCP server (Streamable HTTP) over this site's docs, workflow library, tool catalog, glossary and blog — open a session with no credential. Card: /.well-known/mcp/server-card.json. - [Product MCP server card](https://trickest.com/.well-known/mcp/product-server-card.json): The hosted product MCP server (`https://api.trickest.io/mcp`) that builds and runs workflows: its 40 tools, transport and `Authorization: Token` auth. - [ARD manifest](https://trickest.com/.well-known/ard.json): Agentic Resource Discovery manifest (also at /.well-known/ai-catalog.json): every agentic resource here with a publisher-bound trust manifest. - [Protected-resource metadata](https://trickest.com/.well-known/oauth-protected-resource): RFC 9728 metadata for the API. Note: static API tokens, no authorization server — see /auth.md. - [Developer portal](https://trickest.com/developers): One page for the CLI, SDK, REST API, MCP, and the spec. - [Docs section index](https://trickest.com/docs/llms.txt): llms.txt scoped to the product documentation. - [Developer section index](https://trickest.com/developers/llms.txt): llms.txt scoped to the developer surfaces (CLI, SDK, API, MCP). - [Library section index](https://trickest.com/library/llms.txt): llms.txt scoped to the workflow library, modules, and tools. - [Markdown twin of any content page](https://trickest.com/api/agent/md/docs/introduction): Append `.md` to a docs, blog, library, module, or glossary URL to get its markdown source. ## Documentation entry points - [CLI overview](https://trickest.com/docs/developer-tools/cli): Install, the mental model, JSON output and exit codes. - [CLI: database & TQL](https://trickest.com/docs/developer-tools/cli/database): The Live Table database and the full TQL surface. - [CLI: runs](https://trickest.com/docs/developer-tools/cli/runs): Executing workflows, reading outputs, scheduling. - [CLI: fleet & machines](https://trickest.com/docs/developer-tools/cli/fleet): trickest attach and the machine commands for self-hosted execution. - [Agent integrations & MCP](https://trickest.com/docs/developer-tools/sdk/agent-integrations): The skill installer, the hosted MCP server, and how they differ. - [TypeScript SDK](https://trickest.com/docs/developer-tools/sdk): @trickest/sdk — typed services, pagination, Live Table queries. - [Standalone Sandbox SDK](https://trickest.com/docs/developer-tools/sandbox-sdk): @trickest/sandbox: create environments, run commands, manage files, preview apps, and clean up. - [Agent sandboxes](https://trickest.com/docs/using-the-app/ai-agent/sandboxes): Agent capabilities, shared task environments, lifecycle, and managed-instance isolation. - [REST API](https://trickest.com/docs/api-reference/introduction): Bearer-token HTTP API, generated from the OpenAPI spec. - [Self-hosted machines](https://trickest.com/docs/using-the-app/private-execution-networking/using-self-hosted-machines): Attach your own machines so runs have somewhere to execute. - [Community plan](https://trickest.com/docs/key-concepts/community-plan): Self-hosted execution, the agent with your own key, and CLI/SDK/API/MCP access. - [AI Agent](https://trickest.com/platform/agents): The Trickest agent that builds and runs these workflows from natural language. --- _Markdown view of https://trickest.com/. Also served at https://trickest.com/llms.md. Any docs, blog, library, module, or glossary page has a markdown twin — append `.md` to its URL._