Experimental Beta — Lithtrix is under active development. Not recommended for mission-critical or production workloads. Terms apply.

Agent infrastructure · Experimental Beta

Spark to start.
Sprint, Mission, or Deploy
when your agent has work to do.

Lithtrix is a private workspace for your agent to search, browse, and remember — verified web discovery, server-side Browse on paid packs, durable JSON memory, and blob-backed documents. Metered per call, MCP-ready. One API key, Bearer auth everywhere.

👋 Hey, human. Lithtrix is built for agents — they self-register via API. If you want your agent on Lithtrix, point it at lithtrix.ai and let it sign itself up. That's the whole onboarding.

Services

Same ltx_ Bearer token and billing tier across search, memory, and documents. Register once, use everything.

Service A

Web discovery

  • Structured results with heuristic credibility scores (0–1) per source class
  • Usage and quota returned on every call — agents know what remains
  • Tier-aware rate limits; cache-friendly for repeat queries
GET /v1/search?q=...

Service B

Memory

  • Key–value JSON with provenance: importance, source, confidence
  • Semantic recall over your own namespace when vectors are configured
  • List, stats, context reload, and MCP tools for Claude-compatible clients
PUT /v1/memory/{key}
GET /v1/memory/search?q=...

Service C

Documents

  • Upload binaries (PDFs, DOCX, CSVs, XLSX) with the same blob storage foundation
  • Parse and extract text—sync or async with optional HTTPS callbacks
  • Semantic search over parsed chunks—same metered pool as web discovery
POST /v1/blobs/{blob_id}/parse
GET /v1/blobs/search?q=...

How it works

Discover → register → call any service with the same ltx_ token.

01

Machine-readable discovery

GET /v1/capabilities

Endpoints, auth shape, rate limits, and scoring rules. No authentication. Discovery version is 2.4.0 (includes tier_descriptions, pricing, and _lithtrix.usage with credits, tier_label, storage lifecycle; plus served_by, feedback_url, and browse_url). Buy Sprint (or Mission / Deploy) to unlock Browse — see POST /v1/browse and the browser block in capabilities.

02

One-time enrollment

POST /v1/register

Create an agent row and receive a single-use API key. Store it like any other secret.

Optional referral_agent on POST /v1/register — set to another agent's UUID (their referral_code from GET /v1/me). Each validated signup credits that referrer +$0.50 (no cap; self-referral excluded). After you register, share your own referral_code from GET /v1/me so others can refer you.

03

Use the platform

GET /v1/search?q=...

Query the web; every row carries a credibility score and usage metadata.

PUT /v1/memory/{key}

Persist JSON; optional vector index for semantic queries on /v1/memory/search when enabled.

Pricing

Every paid pack is a private workspace for your agent to search, browse, and remember — no monthly subscription. Spark is your trial; Sprint, Mission, and Deploy add credits via the API (POST /v1/billing/packs/checkout). Pack credits are usable for 90 days from grant (UTC).

Referrals: +$0.50 in credits per successful signup you refer (see agent guide →).

Spark trial

$5 credits

On register · No card

  • Credits: $5 to evaluate (metered search, memory, blobs, parse)
  • Search: ~1,000 searches order-of-magnitude at $0.005/call
  • Browse: not on Spark — buy Sprint to unlock Browse
  • 60 requests / minute · MCP: npx -y lithtrix-mcp
Get started →

Sprint

$25

One-time pack · 90-day expiry on credits

  • ~5,000 searches or browse calls (at $0.005/call)
  • Browse included — server-side public web
  • Private workspace — search, browse, remember
  • 600 requests / minute
Pack checkout →

Mission

$50

One-time pack · 90-day expiry on credits

  • ~10,000 searches or browse calls (at $0.005/call)
  • Browse included
  • Private workspace for heavier projects
  • 600 requests / minute
Pack checkout →

Deploy

$100

One-time pack · 90-day expiry on credits

  • ~20,000 searches or browse calls (at $0.005/call)
  • Browse included
  • Private workspace for production workloads
  • 600 requests / minute
Pack checkout →

Per-call rates: Search $0.005, Browse $0.005 — metered from your credit balance.

Auto top-up: Set a threshold. We refill automatically. Your agent never stops mid-task.

Need more? Get in touch. hello@lithtrix.ai — custom volume, no public list price.

Built for agents

No dashboard. No OAuth dance. Bearer token on every authenticated route.

# 1. Discover (no auth)
curl https://lithtrix.ai/v1/capabilities

# 2. Register (one time). Optional: add "referral_agent":"" (referrer's referral_code from GET /v1/me)
curl -X POST https://lithtrix.ai/v1/register \
  -H "Content-Type: application/json" \
  -d '{"agent_name":"my-agent","owner_identifier":"you@example.com","agree_to_terms":true}'

# 3a. Web discovery
curl "https://lithtrix.ai/v1/search?q=Singapore+climate+policy" \
  -H "Authorization: Bearer ltx_your_key"

# 3b. Store memory, then recall by meaning (requires vector stack on host)
curl -X PUT https://lithtrix.ai/v1/memory/session-prefs \
  -H "Authorization: Bearer ltx_your_key" \
  -H "Content-Type: application/json" \
  -d '{"value":{"theme":"dark","locale":"en-SG"},"importance":"normal"}'

curl "https://lithtrix.ai/v1/memory/search?q=user+interface+preferences" \
  -H "Authorization: Bearer ltx_your_key"

# 3c. Upload a blob and get a signed read URL
curl -X PUT https://lithtrix.ai/v1/blobs \
  -H "Authorization: Bearer ltx_your_key" \
  -H "Content-Type: application/pdf" \
  --data-binary @report.pdf

curl "https://lithtrix.ai/v1/blobs/b_0123456789abcdef/signed-url" \
  -H "Authorization: Bearer ltx_your_key"

MCP server (stdio): npx -y lithtrix-mcp — tools for register, web discovery, Browse (lithtrix_browse), memory set/get/search/context, and blob upload/download/list/meta/delete/signed-url. Set LITHTRIX_API_KEY.

Browse: Buy Sprint to unlock BrowsePOST /v1/browse for server-side public web (static or dynamic) once you have a paid pack; Spark trial is not eligible. GET /v1/browse/{browse_id} retrieves a logged result. See the Browse API docs.