API Reference

store_memory

Store a new memory with automatic categorization and tagging.

Parameters

Example

store_memory({
  content: `# JWT Refresh Token Implementation

## Architecture
- Access tokens: 15min expiry
- Refresh tokens: 7 days, HTTP-only cookies
- Storage: Redis for refresh token registry

## Files Modified
- src/auth/jwt.ts
- src/middleware/authMiddleware.ts`,

  project_name: "payment-service",
  memory_type: "code-snippet",
  related_files: [
    "src/auth/jwt.ts",
    "src/middleware/authMiddleware.ts"
  ],
  tags: ["authentication", "jwt", "security"],
  importance_score: 0.9
})

// Returns: { memory_id: 1234 }

search_memories

Search stored memories using semantic or keyword search.

Parameters

Search Modes

Example

search_memories({
  query: "authentication implementation decisions",
  project_name: "payment-service",
  search_mode: "semantic",
  memory_type: "decision"
})

// Returns summaries - use expand_memories for full content

File Path Patterns

// Exact file
file_path: "src/auth/jwt.ts"

// All files in directory
file_path: "src/auth/%"

// Any file containing 'jwt'
file_path: "%jwt%"

// All TypeScript files in src
file_path: "src/%.ts"

expand_memories

Get full content of memories after search returns summaries.

Parameters

*At least one of memory_ids or chunk_ids required

Example

expand_memories({
  memory_ids: [1234, 1245, 1289]
})

// Returns full content for each memory

get_insights

Analyze patterns, trends, and learnings across your work.

Parameters

Analysis Types

• comprehensive: Full overview with themes and action items
• patterns: Recurring approaches and design patterns
• quality: Bug patterns, root causes, recurring issues
• learning: Knowledge evolution and growth
• progress: Milestones and project momentum
• technical_debt: Cleanup candidates and refactoring needs

Example

get_insights({
  analysis_type: "quality",
  topic: "authentication bugs",
  project_name: "payment-service"
})

// Returns root cause analysis, fix approaches,
// recurrence patterns, recommendations

start_thinking / add_thought

Structured thinking sequences for complex decisions.

start_thinking Parameters

add_thought Parameters

Example

// Start sequence
start_thinking({
  goal: "Evaluate GraphQL migration strategy",
  project_name: "payment-service"
})
// Returns: { sequence_id: "seq_abc123" }

// Add observations
add_thought({
  sequence_id: "seq_abc123",
  thought_type: "observation",
  thought: "50+ REST endpoints, significant migration effort"
})

// Create branches for alternatives
add_thought({
  sequence_id: "seq_abc123",
  thought_type: "branch",
  branch_name: "Big Bang Migration",
  thought: "Replace all REST at once..."
})

// Conclude (saves as decision memory)
add_thought({
  sequence_id: "seq_abc123",
  thought_type: "conclusion",
  thought: "Recommend gradual migration because..."
})

tasks

Create, complete, and manage tasks with memory linking.

Parameters

Example

// Create task
tasks({
  action: "create",
  task_description: "Add rate limiting to login endpoint",
  project_name: "payment-service"
})
// Returns: { task_id: "task_xyz789" }

// Link related memories
tasks({
  action: "link_memories",
  task_ids: ["task_xyz789"],
  memory_ids: [1234, 1245],
  project_name: "payment-service"
})

// View pending tasks
tasks({
  action: "get",
  project_name: "payment-service",
  include_completed: false
})

// Complete task
tasks({
  action: "complete",
  task_ids: ["task_xyz789"],
  project_name: "payment-service"
})

More Tools

Additional tools available in Recallium:

• recallium — Magic summon to load all project context at once
• get_rules — Get agent behavioral rules and guardrails
• session_recap — Get summary of recent activity
• modify_memory — Update, inactivate, or reactivate memories
• projects — Manage project briefs, PRDs, and implementation plans

For complete documentation, see the GitHub repository.