PRODUCT

For Coding Agents For Agentic Products For Edge Devices For Inference Clusters

Developer API

One API call to deploy a background agent.

Set your instructions, connect your tools, and let the agent handle the rest. Long context, complete observability, and usage-based pricing that scales.

Start Building Free →Read the Docs

Why Subconscious

Built for agents that do challenging work.

Long context

Millions of tokens per run

Our inference runtime prunes context on the GPU at runtime. Your agents handle massive tasks without losing accuracy or blowing up costs.

Simple to configure

A prompt and your tools

No workflow DAGs, no orchestration code. Write detailed instructions, attach your tools, and the model figures out the rest.

Fast & inexpensive

Each token charged once

GPU-native context pruning means tokens are never re-sent. You get 10x cheaper runs at scale compared to framework-based agents.

How it works

Three steps. No orchestration code.

Step 01

Set your instructions

Write a detailed prompt describing what your agent should do. The model interprets goals, constraints, and output format natively. These can be long, hundred line instruction sets!

const run = await client.run({
  engine: 'tim',
  input: {
    instructions: `Research the top 5 AI papers
published this week. For each, summarize the
key contribution, methodology, and potential
real-world applications. Output as markdown.`,

Step 02

Connect your tools

Attach hosted platform tools, authenticated third-party integrations, or your own custom endpoints. Mix and match freely.

    tools: [
      { type: 'platform', id: 'web_search' },
      { type: 'mcp', server: 'https://your-mcp-server.com' },
      { type: 'platform', id: 'my_custom_tool' }
    ],

Step 03

Kick off an agent

Fire a single API call. The agent reasons, calls tools, manages its context, and returns results. Sync or async, your choice.

    options: { awaitCompletion: true }
  }
});

console.log(run.result?.answer);

Observability

Complete visibility into every decision.

Every run exposes a full trace of the agent's reasoning, tool calls, and results. See exactly why it made each decision, what data it used, and where it self-corrected.

✓Structured reasoning trace for every step
✓Full tool call inputs and outputs
✓Context window usage and pruning events
✓Token-level cost breakdown per run

THINKThe user wants recent AI papers. I'll search the web first, then scrape the top results for details.

TOOLweb_search("top AI papers published this week 2026")

RESULT5 results returned. Scraping arxiv.org/abs/2602.1847...

THINKFirst paper covers a new attention mechanism. Let me get the remaining 4 and compile the summary.

TOOLweb_scrape("https://arxiv.org/abs/2602.1923")

PRUNEContext at 82%. Pruning search metadata, keeping paper summaries.

Try it now

Get started on the Playground

Test prompts, attach tools, and see full reasoning traces in real time. No code required, just type and run.

Open Playground →

Tools

Use ours, connect yours, or mix both.

Agents need tools to take action. Use our hosted tools out of the box, connect authenticated third-party integrations, or point the agent at your own endpoints.

Hosted tools

Search, scrape, and browse the web out of the box. No API keys, no setup.

Web SearchWeb ScrapeDeep ResearchCode Sandbox

Authenticated integrations

Connect third-party services through our partners. We handle OAuth and token refresh.

GmailSlackHubSpotLinearNotionGoogle Drive

Custom tools

Point the agent at any HTTP endpoint, MCP server, or OpenAPI spec. Full control.

Your REST APIMCP ServersOpenAPI SpecsWebhooks

Pricing

Clear limits, usage-based pricing.

Start with $5 in free credits. Pay only for what you use after that. No seat licenses, no hidden fees. Every token is charged once, so costs scale linearly with your workload instead of exponentially.

View Pricing →

$5Free credits to start

10xCheaper at scale vs. frameworks

1 tokenCharged once, never re-sent

Agent Builder

Build your agent without writing code.

Our agent builder walks you through prompt writing, tool selection, and configuration. We'll help you craft effective instructions and connect everything your agent needs to get the job done.

Guided prompt writing

We help you write detailed, effective instructions that get the best results from the model.

Visual tool connection

Browse available tools, configure authentication, and test connections before your agent goes live.

Test before you ship

Run your agent in the builder, inspect the full reasoning trace, and iterate until it works perfectly.

Try the Agent Builder →

Deployment

Run on the platform, via API, or on a trigger.

dashboard

Platform

Run agents directly from the Subconscious dashboard. View results, reasoning, and tool calls in real time.

POST /v1/runs

API

Integrate into your product with a single API call. Sync or async, with webhook delivery for background runs.

cron / webhook

Triggers

Schedule agents on a cron, fire them on a webhook, or chain them together. Set it and forget it.

Templates

Take action in the real world.

Start from a template and customize. Each one is a real-world workflow you can deploy in minutes.

Deep Research Agent

Search the web, read papers, and compile comprehensive research reports.

web_searchweb_scrape

Lead Enrichment

Take a list of leads and enrich each with company data, social profiles, and news.

web_searchcustom_api

Competitor Monitor

Track competitor pricing, features, and announcements on a recurring schedule.

web_searchweb_scrapeemail

Support Ticket Triage

Read incoming tickets, classify priority, draft responses, and route to the right team.

custom_apislack

Content Pipeline

Research a topic, draft a blog post, create social media variants, and schedule publishing.

web_searchcustom_api

Data Extraction

Scrape structured data from websites, PDFs, and documents into your database.

web_scrapecode_sandbox

Get started today.

$5 free credits. No credit card required.

Start Building Free →Read the Docs