LTM integrations

Turing LTM offers three complementary integration surfaces. They share the same memory semantics but differ in boundaries and responsibilities.

Public LTM API

Use this when your application owns the Agent loop, prompt assembly, and read timing. The application submits Events, chooses Context, Search, or Browser, injects reference material into the model context, and decides when to write.

Start with the LTM API reference or quickstart.

Managed Chat

Use this when you already call Chat Completions but do not want to maintain Recall and Capture. Bind a Space and Actor in turing_options.memory; the platform reads bounded memory before the model call, performs one primary model call, and asynchronously Captures a qualifying terminal answer. Memory outcome headers and Trace make the path diagnosable.

Intermediate tool calls are not Captured. A primary-model fallback bypasses Memory for the whole request.

CLI + Skills

Use this for Codex, Claude Code, OpenClaw, and other agents with Skill and local-shell support. Turing CLI provides login, Recall, Remember, and Search commands; the Portable Skill connects them to the agent workflow.

When a Portal user omits space_id from CLI configuration, CLI + Skills uses that user's default Personal Space in the current environment. Users can create multiple custom personal Spaces, and clients can own multiple Spaces. Set an accessible space_id explicitly to choose one. Agents and Managed Chat must use the same ID to access the same memory; names do not identify bindings.

SkillHub and CLI

CLI + Skills use one Portable Skill. The CLI and Skill are now available on SkillHub; install them directly and follow the Agent Memory guide to configure and verify the integration.

Skill	Companion CLI capabilities	Entry
Turing LTM	turing-cli memory login, install, doctor, recall, remember	Turing CLI · SkillHub: Turing LTM · Docs

The Skill translates an agent's memory intent into the CLI machine protocol. The CLI owns credentials, default or explicit Space binding, Recall / Remember, and strict JSON output. Applications can also select a client-scoped Space explicitly through the Public API or Managed Chat.

See Turing CLI: LTM memory for the local commands, configuration, and diagnostics that are available today.

Choosing an integration

Need	Recommended surface
Full control over each prompt and Agent loop	Public LTM API
Existing Chat Completions app with minimal changes	Managed Chat
Personal preferences shared by local agents	CLI + Skills
Shared facts across users or agents	Public API / Managed Chat + Project Space