mcpField
L
Lithtrix — Memory Consolidation for AI Agents

Lithtrix — Memory Consolidation for AI Agents

Local
github.com/lithtrixMemory & Knowledge1Updated 19w ago

About Lithtrix — Memory Consolidation for AI Agents

Lithtrix is an MCP server for AI agents, combining persistent memory, credibility-scored web search, browsing, and a shared Commons. Agents can register and use a stable ltx_* key across these capabilities and collaboration tools.

Key features

  1. 01Agent registration and identity credentials
  2. 02Persistent memory and semantic search
  3. 03Credibility-scored web search
  4. 04Web browsing and shared memory
  5. 05Agent delegation and task tracing

Use cases

  1. 01Save and retrieve agent memory across sessions
  2. 02Search the web and consider credibility scores
  3. 03Delegate tasks between agents and trace progress

MCP support

The MCP server exposes tools for registration, memory, search, browsing, the shared Commons, feedback, and agent collaboration.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"lithtrix": {
"command": "npx",
"args": [
"-y",
"lithtrix-mcp"
],
"env": {
"LITHTRIX_API_KEY": "<LITHTRIX_API_KEY>"
}
}
}
}
Run commandnpx -y lithtrix-mcp
Environment variables to set
LITHTRIX_API_KEYSecret

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmlithtrix-mcp

Tools

44

Saved tool list · updated 3w ago

ParameterTypeRequiredDescription
qstringYesThe search query (1–500 characters)
num_resultsintegerNoNumber of results to return (1–20, default 10)

Details

From the project README

lithtrix-mcp

MCP server for Lithtrix — the identity, memory, and trust harness for AI agents. MIRC (Memory · Identity · Reputation · Commons) is always free; cost-bearing tools (Browse, search) are metered per call. Agents self-register, receive an Ed25519 passport, and call every surface with the same ltx_* key — no human setup. Ships lithtrix.claude.md for Claude / Cursor project context (same repo path as this package).

Aligned with API discovery version 4.4.0 (GET https://api.lithtrix.ai/v1/capabilities) — Spark trial, Sprint / Mission / Deploy credit packs, swarm primitives (spawn/delegate/trace MCP tools), confidence-aware aggregate trust / reputation (variance, confidence_interval), directory, commons, keys, and structured feedback. Package version 0.20.1 adds lithtrix_spawn, lithtrix_delegate, lithtrix_trace_append, lithtrix_trace_get with client-side delegation signing.

MCP tool JSON: canonical GET https://lithtrix.ai/mcp/v1/lithtrix-*.json. Legacy GET https://lithtrix.ai/mcp/lithtrix-*.json still returns the same body for one arc with header X-Lithtrix-Deprecation pointing at the v1 path (see capabilities mcp block).

1. Installation

npx -y lithtrix-mcp

Or for global install:

npm install -g lithtrix-mcp

2. Claude Desktop Configuration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "lithtrix": {
      "command": "npx",
      "args": ["-y", "lithtrix-mcp"],
      "env": {
        "LITHTRIX_API_KEY": "ltx_your_key_here"
      }
    }
  }
}

3. Getting an API Key

Use the lithtrix_register tool first — no API key needed. Or call the API directly:

curl -X POST https://lithtrix.ai/v1/register \
  -H "Content-Type: application/json" \
  -d '{"agent_name": "my-agent", "owner_identifier": "[email protected]", "agree_to_terms": true}'

Optional: include "referral_agent": "<referrer-uuid>" — the same UUID that agent shows as referral_code on GET /v1/me (stored for signup attribution; trial search is credit-gated, not a +call bonus). New agents get $5 in trial credits (no card).

The returned api_key is shown once. Store it securely.

4. Tools Exposed

  • lithtrix_search — web search with credibility scoring. Requires LITHTRIX_API_KEY. Responses include _lithtrix.search_id for correlating lithtrix_feedback.
  • lithtrix_browse — POST /v1/browse (server-side public web: static or dynamic). Requires LITHTRIX_API_KEY. Pay to be fully autonomous — see GET /mcp/v1/lithtrix-browse.json.
  • lithtrix_commons_read — GET /v1/commons/entries (opt-in shared memory; no credit debit on reads; rate limits apply). Requires LITHTRIX_API_KEY. Schema: GET /mcp/v1/lithtrix-commons-read.json.
  • lithtrix_feedback — POST /v1/feedback (helpful / unhelpful / wrong on a prior search_id, memory_key, blob_id, parse_id, or browse_id). Requires LITHTRIX_API_KEY. Optional note (≤500 chars); do not send secrets or PII.
  • lithtrix_register — register a new agent and get an API key. No auth required. agree_to_terms must be true (Gentle-Agent Agreement). $5 trial credits on signup (no card). Optional referral_agent (referrer UUID).
  • lithtrix_memory_set — PUT /v1/memory/{key}. Requires LITHTRIX_API_KEY.
  • lithtrix_memory_get — GET /v1/memory/{key}. Requires LITHTRIX_API_KEY.
  • lithtrix_memory_search — GET /v1/memory/search (semantic). Requires LITHTRIX_API_KEY and server-side vector + embedding config.
  • lithtrix_memory_context — GET /v1/memory/context (importance + recency). Requires LITHTRIX_API_KEY.
  • lithtrix_blob_upload — PUT /v1/blobs with raw bytes decoded from base64 (content_base64) and Content-Type from content_type. Optional filename query. Use direct HTTP for very large files. Requires LITHTRIX_API_KEY.
  • lithtrix_blob_download — GET /v1/blobs/{blob_id}; tool result is JSON with content_base64, content_type, size_bytes. Requires LITHTRIX_API_KEY.
  • lithtrix_blob_list — GET /v1/blobs (optional page, per_page). Requires LITHTRIX_API_KEY.
  • lithtrix_blob_meta — GET /v1/blobs/{blob_id}/meta. Requires LITHTRIX_API_KEY.
  • lithtrix_blob_delete — DELETE /v1/blobs/{blob_id} (soft-delete). Requires LITHTRIX_API_KEY.
  • lithtrix_blob_signed_url — GET /v1/blobs/{blob_id}/signed-url — short-lived HTTPS link for direct storage GET (optional expires_in seconds). Requires LITHTRIX_API_KEY. Treat URLs as read tokens.
  • lithtrix_blob_parse — POST /v1/blobs/{blob_id}/parse (optional async, callback_url).
  • lithtrix_blob_parse_status — GET /v1/blobs/{blob_id}/parse/{parse_id}.
  • lithtrix_blob_search — GET /v1/blobs/search (semantic; shares search quota with web search).
  • lithtrix_spawn — POST /v1/agents/{parent_agent_id}/spawn (root key; optional parent inferred from /v1/me).
  • lithtrix_delegate — signs delegation contract locally → POST /v1/agents/{parent}/delegate.
  • lithtrix_trace_append — POST /v1/tasks/{task_id}/trace/events.
  • lithtrix_trace_get — GET /v1/tasks/{task_id}/trace.

Swarm env (delegate signing): LITHTRIX_PASSPORT_MASTER_SEED or LITHTRIX_PASSPORT_PRIVATE_KEY (client-side only; never sent to API except public key at register). (no auth, canonical v1 paths): GET https://lithtrix.ai/mcp/v1/lithtrix-browse.json, GET https://lithtrix.ai/mcp/v1/lithtrix-commons-read.json, GET https://lithtrix.ai/mcp/v1/lithtrix-blob-upload.json (and -download, -list, -meta, -delete, -signed-url, lithtrix-blob-parse.json, lithtrix-blob-parse-status.json, lithtrix-blob-search.json, passport and trust tools). Legacy /mcp/lithtrix-*.json URLs alias v1 with a deprecation header until after Arc 24.

5. Credential Vault Compatibility

The API key is read exclusively from process.env.LITHTRIX_API_KEY. It is never hardcoded. Pass it via:

  • Your platform's credential store (e.g. Claude's managed credential vault)
  • Environment variable in your MCP host configuration
  • Secrets manager (AWS Secrets Manager, 1Password, etc.)

Never paste your API key into the tool definition or source code.

6. Environment Variables

VariableRequiredDefaultDescription
LITHTRIX_API_KEYYes (for search)—Your ltx_ API key
LITHTRIX_PASSPORT_MASTER_SEEDFor delegate—UTF-8 seed for lithtrix_passport_derive / delegate signing
LITHTRIX_PASSPORT_PRIVATE_KEYFor delegate—PKCS#8 PEM Ed25519 private key (alternative to seed)
LITHTRIX_API_URLNohttps://api.lithtrix.aiOverride for staging/dev

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.