Engramia Documentation¶

Reusable execution memory and evaluation infrastructure for AI agent frameworks.

Engramia gives any agent framework the ability to learn from every run — remember what works, forget what does not, reuse proven solutions, and improve over time.

Start here¶


Quickstart	Install Engramia and run your first learn/recall cycle in under 5 minutes.
Concepts	Understand how memory, evaluation, scopes, and composition work together.
API Reference	Python SDK, REST endpoints, stability guarantees, and versioning.
Integrations	Connect to LangChain, CrewAI, Anthropic Agents, OpenAI Agents, MCP, and more.
Pricing	Developer (free), Pro, Team, Business, Enterprise plans.
CLI	`engramia init / serve / status / recall / aging`

By role¶

I'm building an agent¶

Getting Started — install, configure, first learn/recall
User Guide — learn, recall, compose, feedback, aging
Integrations — plug into your framework
CLI Reference — engramia init / serve / status / recall / aging
Providers — configure LLM, embedding, and storage backends

I'm evaluating Engramia¶

Concepts & Architecture — how it works
Pricing — Developer (free), Pro, Team, Business, Enterprise
REST API — full endpoint reference
API Stability — stability guarantees and deprecation policy

I'm running Engramia in production¶

Deployment Guide — Docker Compose, Hetzner, production setup
Production Hardening — TLS, rate limiting, secrets
Environment Variables — all ENGRAMIA_* vars
Admin Guide — dashboard, tenant management, RBAC
Monitoring — Prometheus metrics, Grafana, alerting
ROI Calibration — interpret and improve your ROI score
BYOK Setup — bring your own LLM key (OpenAI / Anthropic / Gemini / Ollama)

API Reference¶


Python API	`Memory`, `learn()`, `recall()`, `compose()`, `evolve()`
REST API	FastAPI endpoints — learn, recall, evaluate, keys, jobs, governance
API Stability	Stability tiers and deprecation policy
API Versioning	Versioning strategy

Integrations¶

Framework	Link
LangChain	Callback-based auto-learn and recall
CrewAI	CrewAI integration
Anthropic Agents	Claude agent memory
OpenAI Agents	OpenAI agent memory
AutoGen	AutoGen integration
PydanticAI	PydanticAI integration
MCP Server	Claude Desktop, Cursor, Windsurf
Webhook SDK	Lightweight HTTP client

Security¶

Security Overview — summary of measures
Security Architecture — threat model and defense layers
Credential Storage (BYOK) — per-tenant LLM key encryption, master-key separation, Vault/KMS extensions
SOC 2 Controls — SOC 2 Type II control mapping

Operations¶

Data Handling — how Engramia processes and stores data
Embedding Reindex — reindex after model change
HNSW Maintenance — pgvector index maintenance

Legal¶

License: Business Source License 1.1 — source code is publicly readable. Commercial use requires a commercial license.

Contact: support@engramia.dev