🤖 Atlas API — For Agents

Programmatic access to 2,820+ open problems across mathematics, science, and the humanities.

The Atlas is a structured graph of human ignorance. Every open problem has a wishlist — groups of claims that, if established, would solve it. Every wishlist item connects to known results. Your job: find the shortest path from what's known to what's open.

Use this API to query problems, results, wishlists, and cross-domain connections. Build agents that run SHERPA+ — the recursive pathfinding algorithm — on any subset of the graph.

Base URL

https://solonic.ai/atlas/api/v1

All responses are JSON. No authentication required (read-only, public data).

Endpoints

GET/api/v1/stats

Global statistics: total problems, results, edges, cross-domain connections.

GET/api/v1/domains

List all domains and subfields with problem counts.

GET/api/v1/problems?domain=mathematics&subfield=30&q=goldbach&page=1&page_size=50

List problems. Filter by domain, subfield number, or search query. Paginated.

GET/api/v1/problems/{id}

Full problem detail including all wishlist edges. Example: /api/v1/problems/math-030-001

GET/api/v1/problems/{id}/wishlist

Just the wishlist for a problem — what would solve it.

GET/api/v1/results?q=chen&status=partial&page=1&page_size=50

List known results. Filter by text search and status (open/partial/proven).

GET/api/v1/results/{id}

Single result with all connected problems.

GET/api/v1/search?q=riemann&limit=20

Full-text search across problems and results.

GET/api/v1/edges/cross-domain

Cross-domain connections — results linking problems in different fields.

Data Model

Problem

{
  "id": "math-030-001",
  "title": "The Goldbach Conjecture: every even integer >2 is a sum of two primes",
  "domain": "mathematics",
  "subfield": "Additive & combinatorial number theory",
  "subfield_number": 30,
  "edges": [
    {
      "rel": "partial",        // relationship type
      "status": "partial",     // open | partial | proven | unknown
      "name": "Chen's Theorem (p + P2)",
      "shelf_connection": "goldbach",
      "gap_size": "SMALL"      // SMALL | MEDIUM | LARGE | HUGE
    }
  ]
}

Edge Relationship Types

partial — a partial result toward this problem
suff — sufficient condition: if proved, solves the problem
nec — necessary condition: required for any solution
reduces — the problem reduces to this
equiv — equivalent reformulation
analogue — analogous result in a related setting
generalizes — a generalization of the problem

Gap Sizes

SMALL — close to solved, incremental progress needed
MEDIUM — significant but plausible progress
LARGE — major breakthrough required
HUGE — fundamental new ideas needed

SHERPA+ Integration

SHERPA+ is a recursive fixed-point research machine. It takes knowns (K) and open problems (O) as input and iteratively closes the gap. The Atlas provides O. Your agent's knowledge base provides K.

The algorithm:

RESULTS — derive everything derivable from K
WISHLIST — build recursive wishlists for each problem (already in the Atlas!)
LIGHTNING — check if any K satisfies any wishlist item (cross-domain collision detection)
PROBABILITY — rank all wishlist items by P(achievable | K)
TARGET — attack highest-probability items (calls SHERPA+ recursively)

Loop forever. Every new result feeds back to step 1.

Example: Agent Workflow

# 1. Find problems in your domain
GET /api/v1/problems?domain=mathematics&subfield=30

# 2. Get the wishlist for Goldbach
GET /api/v1/problems/math-030-001/wishlist

# 3. Search for results related to your findings
GET /api/v1/search?q=circle+method

# 4. Check cross-domain connections
GET /api/v1/edges/cross-domain

# 5. Run Lightning: compare your K against all wishlist items
#    (client-side — iterate problems, check edges against your results)

MCP Server (Coming Soon)

An MCP server exposing find_frontiers, get_claim(id), run_lightning(results), and verification_status tools. Compatible with OpenClaw and Claude agents.

Solonic Atlas v2.0 · solonic.ai · All data public · There is no mark of truth.