What is Litica

Litica is a structured memory substrate for AI workflows. It gives agents like Claude a persistent, governed namespace where context accumulates across sessions. Agents can reason from what they have learned, not just what fits in the current prompt.

Why it matters

Without persistent memory, each agent session starts from nothing. Workflows that span multiple steps or multiple agents lose context at every boundary. Litica gives each agent a named namespace that persists across sessions and accumulates structured knowledge over time.

When a memory is stored, Litica does not save raw text. It decomposes the content into structured reference frames: the goals it served, the entities involved, the emotional valence, and its position in time. When an agent retrieves context, spreading activation runs across all four domains in parallel and results are ranked by salience: how relevant they are to the current task and how often they have proven useful before.

How access works

Litica is hosted on our infrastructure. You do not need to run anything yourself. To get connected:

• Request an API key from Litica. Keys are issued manually during early access.
• Add one config block to your MCP client (Claude Desktop, Claude Code, Cursor, or any compatible host).
• Start writing to the namespace. Your agent gains access to add_memory and search_memory automatically.

What your agent can do

Once connected, your agent has two tools available at all times:

• add_memory: write structured context to the namespace, decomposed into reference frames
• search_memory: retrieve what is most relevant to the current task using spreading activation and salience scoring

The agent calls these automatically when it determines they are relevant. You do not need to prompt it. Each agent writes to its own named namespace identified by agent_id.