Infinite Context (ICX) Developer Portal
Connect LLMs directly to millions of tokens of persistent associative memory with $O(1)$ recall latency, zero context window replay costs, and real-time auto-sync from Git, Notion, Google Drive, Linear, and Slack.
1. Overview & How It Works
#Standard LLM APIs force applications to re-send entire documents, conversation histories, and codebases in every single API call. This creates quadratic latency, massive token bills, and "needle-in-a-haystack" attention decay beyond 128k tokens.
Infinite Context (ICX) acts as an intelligent memory proxy between your application and any LLM (Claude 3.5, GPT-4o, Gemini 1.5/2.0, DeepSeek, or open weights). When you upload documents or stream chat turns, ICX crystallizes knowledge into a persistent associative memory lattice. During query time, ICX retrieves the exact relevant sub-graphs in constant $O(1)$ time and packs only the required context viewport into your prompt.
Traditional Token Replay
- Re-upload 500k–2M tokens on every turn
- $2.50–$15.00+ per query prompt costs
- Attention dilution and prompt truncation
- Static memory: changes require full re-indexing
Infinite Context (ICX)
- Persistent memory stays in the lattice ($O(1)$ recall)
- Tiny prompt payloads: pay only for active response tokens
- 96.28% exact factual accuracy across 10M+ tokens
- Living auto-sync: delta patches update memory in <2ms
2. 3-Minute Quickstart
#
Get started with ICX in your preferred language or workflow. Autonomous agents and developers can install the official Python SDK via PyPI (pip install calera-agent-memory), drop into LangChain/CrewAI agent swarms, connect via Cursor/Claude MCP, or use the standard OpenAI-compatible base URL (https://icx.api.caleralabs.com/v1).
# pip install calera-agent-memory
from calera_agent_memory import CaleraMemoryClient
# 1. Self-provision a free 1,000,000-node persistent vault (Delaware UETA § 14)
client = CaleraMemoryClient()
vault_info = client.claim_free_vault("developer@company.com", agent_name="my-researcher")
print(f"Provisioned Vault ID: {client.vault_id}")
# 2. Store structured facts or execution state into 4D topological lattice
client.store("NVDA_FY26_Rev", "Estimated at $168B on Blackwell datacenter expansion")
# 3. Sub-5ms topological associative recall with zero context rot
results = client.query("NVDA revenue projection")
print(results)
# pip install calera-agent-memory langchain
from calera_agent_memory import CaleraMemorySubstrate
# Auto-provisions a free 1M-node persistent memory vault under Delaware UETA § 14
memory = CaleraMemorySubstrate(principal_email="developer@company.com")
# Drop directly into LangChain, CrewAI, or AutoGen agent loops
memory.save_context({"input": "Q3 Target Launch"}, {"output": "November 14, 2026"})
context = memory.load_memory_variables({"input": "When is the launch?"})
print(context["history"])
import os
from openai import OpenAI
# Connect standard OpenAI client to ICX API gateway (Zero custom packages required)
client = OpenAI(
api_key=os.environ.get("ICX_API_KEY", "icx_live_your_api_key"),
base_url="https://icx.api.caleralabs.com/v1",
default_headers={
"X-Space-ID": "enterprise_core"
}
)
# Query persistent memory — ICX handles associative memory recall automatically
response = client.chat.completions.create(
model="calera-icx-v1",
messages=[
{"role": "user", "content": "Summarize key risk factors in Section 1A from our ingested filings."}
]
)
print(response.choices[0].message.content)
// Add to .cursor/mcp.json or claude_desktop_config.json
{
"mcpServers": {
"infinite-context": {
"url": "https://icx.caleralabs.com/mcp",
"headers": {
"X-License-Key": "clabs_live_YOUR_KEY",
"X-Space-ID": "enterprise_core"
}
}
}
}
# Ingest text into persistent memory
curl -X POST https://icx.api.caleralabs.com/v1/ingest/text \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"space_id": "knowledge_base",
"text": "Project Apollo Launch Date: November 14, 2026. Target Orbit: 420km LEO. Lead Architect: Dr. Vance.",
"source_id": "apollo_spec_v1"
}'
# Query via standard chat completions
curl -X POST https://icx.api.caleralabs.com/v1/chat/completions \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "X-Space-ID: knowledge_base" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-4o",
"messages": [{"role": "user", "content": "When is Project Apollo scheduled to launch?"}]
}'
3. Authentication & Headers
#
Authenticate every request using your ICX API Key (icx_live_... or icx_test_...). Pass it either as a standard Bearer token or via the X-License-Key header.
| Header | Required | Description |
|---|---|---|
Authorization |
Yes | Standard Bearer token header: Bearer icx_live_... |
X-License-Key |
Optional | Alternative key header (e.g. icx_live_...). Useful for clients that enforce custom auth headers. |
X-Space-ID |
Optional | Memory Space identifier (e.g. corp_docs, repo_main). Default: default. |
X-Session-ID |
Optional | Conversation turn session ID (e.g. sess_dev_102). Tracks multi-turn dialogue state. |
X-BYOK-Provider |
Optional | Bring Your Own Key LLM provider: openai, anthropic, google, or groq. |
X-BYOK-Key |
Optional | Your direct API key for the underlying LLM provider when using BYOK mode. |
4. Chat Completions (/v1/chat/completions)
#
POST /v1/chat/completions is fully compatible with the OpenAI API specification. When a query is sent, ICX traverses your crystallized memory space, retrieves exact factual subgraphs in under 5ms, and injects the ground-truth viewport before dispatching to the LLM.
curl -X POST https://icx.api.caleralabs.com/v1/chat/completions \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "X-Space-ID: production_kb" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-3-5-sonnet-20241022",
"messages": [
{"role": "system", "content": "You are a helpful engineering assistant."},
{"role": "user", "content": "Explain our authentication architecture from the security whitepaper."}
],
"stream": true,
"temperature": 0.2
}'
{
"id": "chatcmpl-icx-8f81a3d0",
"object": "chat.completion",
"created": 1756141200,
"model": "claude-3-5-sonnet-20241022",
"choices": [
{
"index": 0,
"message": {
"role": "assistant",
"content": "According to Section 3 of the security whitepaper, the authentication architecture uses..."
},
"finish_reason": "stop"
}
],
"usage": {
"prompt_tokens": 820,
"completion_tokens": 145,
"total_tokens": 965
},
"vln_telemetry": {
"space_id": "production_kb",
"facts_recalled": 18,
"crystallization_nodes": 452000,
"recall_latency_ms": 1.48,
"token_savings_pct": 99.82
}
}
5. File Ingestion API (/v1/ingest/file)
#
Upload files directly to ICX. The server parses text, code, tables, and hierarchical headings, crystallizing knowledge into your memory space. Supported formats: .pdf, .md, .txt, .docx, .csv, .json, .py, .go, .js, .ts, .rs, .cpp.
curl -X POST https://icx.api.caleralabs.com/v1/ingest/file \
-H "Authorization: Bearer icx_live_your_api_key" \
-F "file=@./legal_contract_2026.pdf" \
-F "space_id=legal_contracts" \
-F "source_id=contract_acme_2026"
{
"status": "success",
"space_id": "legal_contracts",
"source_id": "contract_acme_2026",
"filename": "legal_contract_2026.pdf",
"bytes_processed": 482910,
"nodes_crystallized": 1420,
"edges_formed": 4830,
"crystallization_time_ms": 32.4
}
6. Text & Document Ingestion (/v1/ingest/text)
#
For dynamic snippets, database rows, customer tickets, or webhook payloads, send JSON text directly via POST /v1/ingest/text.
curl -X POST https://icx.api.caleralabs.com/v1/ingest/text \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"space_id": "crm_knowledge",
"source_id": "ticket_98214",
"text": "Customer AcmeCorp requested custom SSO domain routing on cluster us-central1-c. Lead Contact: Sarah Chen.",
"metadata": {
"customer": "AcmeCorp",
"priority": "P1",
"timestamp": "2026-08-25T14:30:00Z"
}
}'
7. Bulk Corpus Ingestion (/v1/ingest/bulk)
#
When migrating large documentation archives, git histories, or multi-gigabyte corpora, use POST /v1/ingest/bulk to dispatch an asynchronous background batch ingestion job.
curl -X POST https://icx.api.caleralabs.com/v1/ingest/bulk \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"space_id": "entire_codebase",
"documents": [
{"source_id": "file_1.ts", "text": "export const CONFIG = {...};"},
{"source_id": "file_2.ts", "text": "export function initDatabase() {...};"}
]
}'
8. Continuous Living Auto-Sync
#ICX Living Auto-Sync connects external SaaS platforms directly to your persistent memory space. Whenever a pull request is merged, a Notion doc is updated, or a Google Sheet is edited, changefeeds are crystallized in real time (<2ms CPU delta execution) without needing to re-upload full files.
Codebase Sync
GitHub & GitLab webhook connectors automatically sync diffs, PR comments, and commits upon push.
Docs & Knowledge
Notion, Google Drive, and Linear connectors update specs, tickets, and tables continuously.
Live Conversations
Slack changefeeds sync decision channels (e.g. #decisions, #incidents) into memory.
9. Connectors & Webhooks API
#Manage auto-sync connectors programmatically or register webhook callbacks.
curl -X POST https://icx.api.caleralabs.com/v1/sync/connectors \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"provider": "github",
"space_id": "backend_repo",
"target": "calera-computing/core-engine",
"branch": "main",
"auto_sync": true
}'
curl -X POST https://icx.api.caleralabs.com/v1/sync/webhook/generic \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"space_id": "ci_builds",
"source_id": "build_20260825_981",
"event_type": "deploy_success",
"payload": "Deployed revision rev-981 to cluster-prod-east. All 42 health checks passed."
}'
10. Instant Memory Purge & Revocation
#When a document is deleted, an employee leaves, or an M&A divestiture occurs, ICX unlinks the associated memory nodes in constant $O(1)$ time. Deleted facts immediately become inaccessible to all LLM queries.
curl -X POST https://icx.api.caleralabs.com/v1/sync/purge \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"space_id": "legal_contracts",
"source_id": "contract_acme_2026"
}'
{
"status": "purged",
"space_id": "legal_contracts",
"source_id": "contract_acme_2026",
"nodes_unlinked": 1420,
"purge_latency_ms": 1.12
}
11. Memory Spaces & Multi-Tenancy
#Memory Spaces provide complete cryptographic boundary isolation. Each space has its own independent graph topology, preventing cross-tenant leakage between different projects, teams, or customers.
curl -X GET https://icx.api.caleralabs.com/v1/memory/spaces \
-H "Authorization: Bearer icx_live_your_api_key"
{
"spaces": [
{
"space_id": "backend_repo",
"nodes_count": 842000,
"sources_count": 142,
"last_sync": "2026-08-25T16:20:00Z"
},
{
"space_id": "legal_contracts",
"nodes_count": 215000,
"sources_count": 38,
"last_sync": "2026-08-24T18:10:00Z"
}
]
}
12. Session State & Reset
#
When building interactive chatbots or coding agents, pass the X-Session-ID header to maintain turn-by-turn conversational state. To clear conversational turns while preserving the background memory space, call POST /api/session/reset.
curl -X POST https://icx.api.caleralabs.com/api/session/reset \
-H "Authorization: Bearer icx_live_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"session_id": "sess_user_482"
}'
13. Python SDK & Standard API Integration
#
ICX provides the official Python SDK calera-agent-memory on PyPI for native sub-5ms topological memory, LangChain/CrewAI substrates, and autonomous vault provisioning under Delaware UETA § 14. In addition, ICX supports standard OpenAI SDK drop-in and high-performance REST APIs (httpx, requests).
# pip install calera-agent-memory
from calera_agent_memory import CaleraMemoryClient, CaleraMemorySubstrate
# 1. Claim a free 1,000,000-node persistent vault at runtime
client = CaleraMemoryClient()
client.claim_free_vault("developer@company.com", agent_name="finance-agent")
# 2. Store facts directly into 4D topological lattice (sub-5ms O(1) recall)
client.store("FY26_EBITDA_Target", "Projected at $4.2B with 38% operating margin")
recalled = client.query("EBITDA margin target")
print("Topological Recall:", recalled)
# 3. LangChain / CrewAI persistent substrate drop-in
memory = CaleraMemorySubstrate(principal_email="developer@company.com")
memory.save_context({"input": "What is our Q4 quota?"}, {"output": "$12.5M ARR"})
history = memory.load_memory_variables({"input": "Q4 quota"})
print("Substrate History:", history)
import os
import requests
from openai import OpenAI
API_KEY = os.environ.get("ICX_API_KEY", "icx_live_your_api_key")
BASE_URL = "https://icx.api.caleralabs.com/v1"
SPACE_ID = "engineering"
# 1. Ingest local files or documentation into persistent memory
with open("architecture_spec.md", "rb") as f:
resp = requests.post(
f"{BASE_URL}/ingest/file",
headers={"Authorization": f"Bearer {API_KEY}", "X-Space-ID": SPACE_ID},
files={"file": f}
)
print("Ingested file:", resp.json())
# 2. Query persistent memory with streaming using standard OpenAI SDK
client = OpenAI(
api_key=API_KEY,
base_url=BASE_URL,
default_headers={"X-Space-ID": SPACE_ID}
)
stream = client.chat.completions.create(
model="calera-icx-v1",
messages=[{"role": "user", "content": "How does our caching layer handle cache misses?"}],
stream=True
)
for chunk in stream:
if chunk.choices[0].delta.content:
print(chunk.choices[0].delta.content, end="", flush=True)
14. Developer CLI (icx)
#
The icx CLI tool enables local directory watching and continuous synchronization directly from your terminal or CI/CD pipelines.
# 1. Authenticate CLI
export ICX_API_KEY="icx_live_your_api_key"
# 2. Watch a local folder and sync file changes in real-time
icx watch ./src --space codebase_main
# 3. One-shot push of a documentation folder
icx push ./docs --space knowledge_base
# 4. View current tenant sync status and memory node count
icx status --space codebase_main
# 5. Instantly purge a deleted module from memory
icx purge deprecated_module_v1 --space codebase_main
15. Model Context Protocol (MCP) Server
#
Connect ICX directly to AI developer tools like Cursor, Claude Desktop, Claude Code, Windsurf, and Antigravity using the certified Model Context Protocol endpoint at https://icx.caleralabs.com/mcp.
{
"mcpServers": {
"icx-memory": {
"command": "npx",
"args": ["-y", "@caleralabs/icx-mcp@latest"],
"env": {
"ICX_API_KEY": "icx_live_your_api_key",
"ICX_SPACE_ID": "primary_workspace"
}
}
}
}
| Tool Name | Description |
|---|---|
icx_remember |
Crystallize a text snippet or code fact into persistent memory. |
icx_recall_scoped |
Query facts and relationships within a specific space. |
icx_search_facts |
Search grounded facts using semantic keywords. |
icx_inspect_space |
Inspect total nodes, sources, and active connectors in a space. |
icx_reset_session |
Reset the current conversation session while preserving memory. |
icx_sync_delta |
Push delta changes from a local file directly into memory. |
icx_purge_source |
Unlink and revoke a memory source in constant time. |
16. Node.js, TypeScript & Go
#import OpenAI from 'openai';
const icx = new OpenAI({
apiKey: process.env.ICX_API_KEY,
baseURL: 'https://icx.api.caleralabs.com/v1',
defaultHeaders: {
'X-Space-ID': 'production_app'
}
});
async function main() {
const completion = await icx.chat.completions.create({
model: 'gpt-4o',
messages: [
{ role: 'user', content: 'What are the SLA response times for Tier-1 incidents?' }
]
});
console.log(completion.choices[0].message.content);
}
main();
package main
import (
"bytes"
"encoding/json"
"fmt"
"io"
"net/http"
"os"
)
func main() {
payload, _ := json.Marshal(map[string]any{
"model": "gpt-4o",
"messages": []map[string]string{
{"role": "user", "content": "Summarize project requirements."},
},
})
req, _ := http.NewRequest("POST", "https://icx.api.caleralabs.com/v1/chat/completions", bytes.NewBuffer(payload))
req.Header.Set("Authorization", "Bearer "+os.Getenv("ICX_API_KEY"))
req.Header.Set("X-Space-ID", "core_specs")
req.Header.Set("Content-Type", "application/json")
client := &http.Client{}
resp, err := client.Do(req)
if err != nil {
panic(err)
}
defer resp.Body.Close()
body, _ := io.ReadAll(resp.Body)
fmt.Println(string(body))
}
17. Semantic Memory Search API (/v1/lattice/search)
#
Directly query the crystallized memory graph to inspect stored facts and relationships without generating an LLM response.
curl -X GET "https://icx.api.caleralabs.com/v1/lattice/search?space_id=engineering&q=database+sharding&limit=5" \
-H "Authorization: Bearer icx_live_your_api_key"
{
"query": "database sharding",
"space_id": "engineering",
"total_matches": 3,
"facts": [
{
"fact_id": "f_8921a",
"text": "User database is sharded across 16 PostgreSQL partitions based on tenant UUID hash.",
"source_id": "db_architecture_v2.md",
"relevance_score": 0.96
}
]
}
18. Rate Limits & Tiers
#ICX pricing is based on active crystallized memory capacity rather than repetitive per-token replay charges.
| Tier | Price | Memory Capacity | Connectors | Rate Limit |
|---|---|---|---|---|
| Free | $0 / mo | 1,350,000 Nodes (~2,700 pages) | 1 Auto-Sync Connector | 60 req / min |
| Builder | $29 / mo | 50,000,000 Nodes (~100k pages) | 5 Auto-Sync Connectors | 300 req / min |
| Team | $149 / mo | 500,000,000 Nodes (~1M pages) | Unlimited Connectors | 1,200 req / min |
| Scale | $999 / mo | 5,000,000,000 Nodes (~10M pages) | Unlimited Connectors | 5,000 req / min |
| Enterprise | Custom | 50B+ Nodes / Dedicated Cluster | Unlimited Connectors | Custom SLA |
19. Error Codes Catalog
#Standard HTTP status codes and machine-readable error strings returned by the ICX Gateway.
| Status & Code | Category | Description & Action |
|---|---|---|
401 UnauthorizedERR_AUTH_INVALID_KEY |
Authentication | Missing, malformed, or revoked API key. Verify Authorization: Bearer header. |
404 Not FoundERR_SPACE_NOT_FOUND |
Memory Space | The requested X-Space-ID does not exist or has not been initialized. |
404 Not FoundERR_CONNECTOR_NOT_FOUND |
Auto-Sync | Specified connector ID was not found for this tenant space. |
429 Too Many RequestsERR_RATE_LIMIT_EXCEEDED |
Rate Limiting | Request rate limit exceeded for your current tier. Retry using exponential backoff. |
429 Capacity LimitERR_TIER_NODE_LIMIT_EXCEEDED |
Lattice Quota | Lattice node capacity reached for your tier plan. Upgrade tier in dashboard to expand nodes. |
500 Server ErrorERR_UPSTREAM_PROVIDER_FAILED |
LLM Upstream | The selected BYOK provider (OpenAI, Anthropic) returned an upstream error. |