Migration Guide - Claude Agent SDK to Letta Code SDK

Fresh

Source: Letta Code SDK Migration

Key Differences

The Claude Agent SDK is query-based with ephemeral sessions. The Letta Code SDK uses persistent agents with multiple conversation threads.

Feature Comparison

Feature	Claude Agent SDK	Letta Code SDK
Persistence	Session-based	Agent-based
Memory	Session state only	Persistent memory blocks
Models	Claude only	Multiple options
Conversations	One per session	Multiple per agent

Claude Agent SDK V2 Migration

Function Mapping

Claude V2	Letta Code SDK
unstable_v2_prompt()	prompt()
unstable_v2_createSession()	createSession()
unstable_v2_resumeSession(session_id)	resumeSession(conversationId)

Before (Claude V2)

import { unstable_v2_createSession, unstable_v2_resumeSession } from '@anthropic-ai/claude-code-sdk';

// Create session
const session = await unstable_v2_createSession();
const sessionId = session.sessionId;

// Resume later
const resumed = await unstable_v2_resumeSession(sessionId);

After (Letta Code SDK)

import { createAgent, resumeSession } from '@letta-ai/letta-code-sdk';

// Create agent (persistent)
const session = await createAgent();
const agentId = session.agentId;

// Resume later - the agent persists, just reconnect
const resumed = await resumeSession(agentId);

Key Insight

With Letta, the agent persists -- just use resumeSession(agentId) to continue where you left off. No need to manage session IDs separately.

Claude Agent SDK V1 Migration

Function Mapping

Claude V1	Letta Code SDK
query() (one-shot)	prompt(message)
query() (multi-turn)	createSession().send() + stream()

Before (Claude V1)

import { query } from '@anthropic-ai/claude-code-sdk';

const result = await query('What does this function do?');

After (Letta Code SDK)

import { prompt, createSession } from '@letta-ai/letta-code-sdk';

// One-shot (same as V1 query)
const result = await prompt('What does this function do?');

// Multi-turn (new capability)
const session = await createSession();
const response = await session.send('Explain the auth module');
const followUp = await session.send('What about the middleware?');
await session.close();

Permissions and Tools

Claude SDK

Limited permission control, typically all-or-nothing.

Letta Code SDK

Fine-grained control:

const session = await createAgent({
  // ... agent config
});

// Resume with restricted tools
const restricted = await resumeSession(session.agentId, {
  allowedTools: ['Read', 'Glob', 'Grep'],
  permissionMode: 'plan',
  canUseTool: (tool) => {
    // Custom approval logic
    return tool !== 'Write';
  }
});

Migration Checklist

• [ ] Replace Claude SDK import with @letta-ai/letta-code-sdk
• [ ] Map V1 query() calls to prompt()
• [ ] Map V2 unstable_v2_* calls to their Letta equivalents
• [ ] Replace session ID tracking with agent ID tracking
• [ ] Add memory block configuration where beneficial
• [ ] Update permission/tool configurations
• [ ] Test that agent memory persists across sessions

See Also

• SDK Overview - Full SDK capabilities
• API Reference - Function signatures
• Agents - Agent lifecycle