Quickstart

Go from zero to your first Neural Snapshot in under five minutes. All you need is a NeuralSnap account and an API key.

Base URL

All API requests in this guide use https://api.neuralsnap.ai/v1

Create an account

Head over to neuralsnap.ai/signup and create your free account. A default brain is provisioned automatically — you can start adding snapshots straight away.

The free tier includes 500K tokens/month and 1K searches — enough to store hundreds of snapshots and test every feature.

Get your API key

Navigate to Settings → API Keys and click Generate Key. Copy the key — you'll need it for every request.

Treat your API key like a password. Never commit it to version control or share it publicly. Use environment variables in production.

Set it as an environment variable so the examples below work out of the box:

bash

export NEURALSNAP_API_KEY="ns_live_your_key_here"

Install an SDK (optional)

You can use raw HTTP requests, but our SDKs handle auth, retries, and types for you:

bash

npm install @neuralsnap/sdk
# or
pnpm add @neuralsnap/sdk

Not required for this tutorial — all examples show both SDK and raw HTTP.

Create your first snapshot

A Neural Snapshot is a structured knowledge unit with up to 16 fields. At minimum, you need a name, core (the key insight), and a type.

5 Snapshot Types

Conviction

Something you strongly believe based on evidence

Belief

A working assumption you operate from

Model

A mental framework for understanding something

Rule

A concrete operating principle you follow

Principle

A high-level guideline that informs decisions

Send a POST to /snapshots:

bash

curl -X POST https://api.neuralsnap.ai/v1/snapshots \
  -H "Authorization: Bearer $NEURALSNAP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "name": "Building in public works",
  "core": "Revenue transparency drives trust and engagement, converting followers into customers faster than polished marketing.",
  "type": "Conviction",
  "proof": "After 3 months of sharing monthly revenue reports, newsletter signups increased 4x and DM inquiries doubled.",
  "confidence": 0.9,
  "tags": ["growth", "content", "transparency"]
}'

The API returns the full snapshot object:

json

{
  "id": "snap_a1b2c3d4",
  "name": "Building in public works",
  "core": "Revenue transparency drives trust and engagement...",
  "type": "Conviction",
  "proof": "After 3 months of sharing monthly revenue reports...",
  "confidence": 0.9,
  "significance_score": 0.85,
  "tags": ["growth", "content", "transparency"],
  "source_type": "api",
  "brain_id": "brain_default",
  "created_at": "2026-02-21T04:00:00Z",
  "updated_at": "2026-02-21T04:00:00Z"
}

16-field snapshots

The full snapshot schema includes fields like proof, counter (counter-arguments), context (origin story), and more. See the Snapshots API reference for every field.

Search your knowledge

NeuralSnap uses 768-dimensional semantic vectors to find knowledge by meaning, not keywords. Ask a question in natural language:

bash

curl "https://api.neuralsnap.ai/v1/snapshots/search?q=how+to+grow+an+audience" \
  -H "Authorization: Bearer $NEURALSNAP_API_KEY"

The response includes scored results:

json

{
  "results": [
    {
      "id": "snap_a1b2c3d4",
      "name": "Building in public works",
      "core": "Revenue transparency drives trust and engagement...",
      "type": "Conviction",
      "score": 0.91,
      "confidence": 0.9,
      "tags": ["growth", "content", "transparency"]
    }
  ],
  "query": "how to grow an audience",
  "count": 1,
  "latency_ms": 23
}

Notice: you searched for "how to grow an audience" and it matched "Building in public works" — because the meaning is related, even though the words are different. That's semantic search.

Crystallize raw text

Instead of manually crafting snapshots, drop in unstructured text and let the AI extract structured knowledge. This is the real power of NeuralSnap — turn meetings, articles, and conversations into searchable knowledge automatically.

bash

curl -X POST https://api.neuralsnap.ai/v1/crystallize \
  -H "Authorization: Bearer $NEURALSNAP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "text": "In our Q4 planning meeting we agreed that doubling down on short-form video is the highest-leverage play. TikTok and Reels drive 3x more signups than blog posts, and the production cost is actually lower once the workflow is dialled in. We also decided to kill the podcast — low ROI, high time investment. Sarah pushed back saying podcasts build deeper trust, but the numbers don\'t support it at our current scale.",
  "source_type": "manual"
}'

One paragraph might produce 2–5 snapshots, each fully structured:

json

{
  "snapshots": [
    {
      "name": "Short-form video is highest leverage for growth",
      "type": "Conviction",
      "core": "TikTok and Reels drive 3x more signups than blog posts at lower production cost.",
      "proof": "Q4 planning data comparison: short-form vs blog signup rates.",
      "confidence": 0.92,
      "tags": ["content-strategy", "video", "growth"]
    },
    {
      "name": "Kill low-ROI podcast at current scale",
      "type": "Rule",
      "core": "Podcast has high time investment with low measurable ROI compared to short-form video.",
      "counter": "Sarah argues podcasts build deeper trust — may be valid at larger scale.",
      "confidence": 0.78,
      "tags": ["content-strategy", "podcast", "prioritization"]
    }
  ],
  "token_usage": { "input": 1240, "output": 890 }
}

Meeting transcripts

Crystallization works best with Fathom or Fireflies transcripts. A 30-minute meeting typically produces 5–10 high-quality snapshots. Connect your meeting tool in Connections.

Explore your brain

Once you've added a handful of snapshots, open the Brain Graph to see your knowledge come alive. Snapshots cluster by semantic similarity, and edges form between related ideas (supports, contradicts, elaborates, caused_by, evolved_from).

Click any node to see the full snapshot. The graph reveals patterns you didn't know existed — like how a product decision from last month connects to a hiring conviction from last quarter.

What you just built

Created a snapshot

Structured knowledge your AI can understand

Searched by meaning

768-dimensional semantic vectors, not keywords

Crystallized raw text

AI-extracted structured knowledge from unstructured text

Next steps

→Core Concepts — understand the 4 knowledge layers (Snapshots, Memories, Knowledge, Emotions)
→API Reference — complete endpoint documentation with request/response schemas
→JavaScript SDK / Python SDK — type-safe SDKs with built-in retries and error handling
→MCP Server — give Claude or Cursor persistent memory with zero code

Overview

Concepts