Zero-dependency client libraries, end-to-end type safety, native SSE streaming, and first-class telemetry for Context, Memory, and Skills across TypeScript, Python, and Go.
Stop wrestling with fragmented REST endpoints and brittle JSON schemas. Our SDK unifies context compression, persistent memory, and sandboxed skill calls into one typed interface.
Configure context window pruning, semantic memory retrieval, and sandboxed skill toolchains through one idiomatic client object. Zero cognitive overhead switching between domains.
Stream token chunks, tool execution telemetry, and memory recall status in real time with automatic connection recovery, backpressure handling, and typed event listeners.
Automatic distributed tracing spans injected for every LLM roundtrip, token savings metric, cache hit, and skill execution. Seamless export to Datadog, Honeycomb, or Grafana Tempo.
Test your autonomous multi-turn loops in CI pipelines without paying provider API fees or risking flaky external network calls. Record, replay, and snapshot agent behavior.
A unified TypeScript client that wires up context pruning, persistent recall, and skill execution in minutes.
// npm install @ai-id/sdk
import { AiIdClient } from '@ai-id/sdk';
const ai = new AiIdClient({
apiKey: process.env.AI_ID_API_KEY,
telemetry: { serviceName: 'billing-agent' }
});
// Run complete agent turn with context compression + memory recall
const stream = await ai.agents.runStream({
entityId: 'cust_8829',
prompt: 'Summarize invoice disputes from last quarter and generate refund draft',
context: {
pruneTokens: true,
targetCompression: 0.4
},
memory: {
recallScope: 'hybrid',
threshold: 0.82
},
skills: ['stripe.disputes', 'zendesk.tickets']
});
for await (const event of stream) {
if (event.type === 'token') process.stdout.write(event.delta);
if (event.type === 'skill:complete') console.log(`[Tool Executed: ${event.skill}]`);
}From microservices to edge runtimes, the AI.ID SDK runs anywhere your code does.
Connection pooling, memory-efficient streams, and concurrency controls designed for services processing millions of agent requests daily.
Sub-millisecond cold starts and native Web Streams support tailored specifically for edge runtimes and Lambda functions.
Seamless integration into Go and Python command-line utilities with built-in spinner, terminal streaming, and auth credential helpers.
We are inviting select engineering teams into our private developer preview. Receive developer API tokens, private npm/PyPI registry invites, and direct access to SDK core maintainers.