Integrations · Developer guide

MAGLA Memory for LlamaIndex agents

Add verified cross-session continuity to LlamaIndex agents without shipping a second memory server.

1. Start with preflight

Ask what verified state the project reached before another durable action.

from magla_memory import Memory import os memory=Memory(os.environ["MAGLA_MEMORY_API_KEY"]) work=memory.preflight("continue llamaindex agents",current_context="",limit=8) print(work)

2. Execute, then verify

Run the real tool or workflow. Keep the returned work_receipt; a model statement alone is not project truth.

3. Close the loop

Use record_outcome after evidence-backed progress or record_failure for a bounded failed approach.

Install without another memory server

1. Get a project key

Create a project on the home page. The key is shown once.

2. Python

curl -fsS https://magla-memory.5-180-47-236.sslip.io/sdk/python -o magla_memory.py export MAGLA_MEMORY_API_KEY='mmc1.…'

3. Remote MCP

URL: https://magla-memory.5-180-47-236.sslip.io/mcp Authorization: Bearer <MAGLA_MEMORY_API_KEY>

No local MAGLA Memory MCP server is required.

Reproducible continuity check

Session A

Preflight → perform one bounded action → verify it → record outcome with evidence.

Session B

Start fresh or switch models → call preflight → confirm the verified checkpoint is the resume point.

Public contract

statuspreflightcontextrecord_outcomerecord_failure

The public surface is intentionally limited to five operations.