selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

mira-OSS logo

Self-host mira-OSS

v2026.10.21-2.1.2

AI assistant with continuous long-term memory

ReleaseOpen SourceAGPL-3.0
Updated9 hours ago
LanguagePython
Setup difficultyAdvanced
Stars
481
Port
1993 (web UI and API)
Install with Docker

What is mira-OSS?

MIRA is a self-hosted AI assistant that approximates a continuous digital entity through a single never-ending conversation thread. It manages its own long-term memory by collapsing older conversation segments into first-person summaries that decay based on relevance rather than calendar time. Memories are recalled automatically via semantic similarity, entity graphs, and reranking before each response. MIRA includes a dynamic tool system where tools register at startup and the assistant can enable additional tools on demand. It runs on PostgreSQL with pgvector for embeddings, Valkey for caching, and HashiCorp Vault for secrets. The architecture uses a subcortical layer for pre-response context retrieval and a Peanut Gallery metacognitive observer for post-response review. Installation options include a one-line installer, Docker (with a multi-stage build), and a hosted version at miraos.org. The project is licensed under AGPL-3.0.

Should you self-host mira-OSS?

mira-OSS scores 49/100 on selfhost.directory (how scoring works) and ranks #47 of 120 in AI Assistant. It has 481 GitHub stars.

It shipped 16 stable releases in the past year. The latest, v2026.10.21-2.1.2, came out on 21 October 2026. New versions land roughly every 8 days.

Setup is advanced: plan for several services and some ongoing maintenance. It is AGPL-3.0-licensed and written mostly in Python.

Its closest alternative on selfhost.directory is hermes-agent (89/100, 257 926 stars). It scores higher overall, so compare the two before you pick. See mira-OSS vs hermes-agent side by side.

Features

  • Autonomous memory decay and recall
  • Bash tool with safety guardrails
  • Continuous conversation thread
  • Domain documents (domaindocs)
  • Dynamic tool system
  • First-person long-term memory
  • Injection screen for external content
  • Multi-user mode with magic-link auth
  • Peanut Gallery metacognitive observer
  • Provider-neutral LLM boundary
  • Self-updating installation
  • Subcortical pre-response context retrieval
  • Terminal chat client (TUI)
  • Text-Based LoRA (behavioral adaptation)

mira-OSS README

How to install mira-OSS

Install mira-OSS with Docker

Environment variables
MIRA_PROVIDER_KEY, MIRA_PROVIDER_NAME, MIRA_PROVIDER_ENDPOINT, MIRA_PROVIDER_MODEL, MIRA_KAGI_KEY, MIRA_SYSTEMONE_API_KEY, MIRA_DB_PASSWORD, MIRA_SUBCORTICAL_ENABLED, MIRA_PEANUTGALLERY_ENABLED, MIRA_PERSONA_ENABLED, MIRA_LOCAL_SESSION_FIRST_NAME, MIRA_CONTEXT_WINDOW_TOKENS, MIRA_COMPACTION_TRIGGER_TOKENS

docker build -t mira-base:latest -f deploy/docker/Dockerfile.base . docker build -t mira:latest -f deploy/docker/Dockerfile .

Interactive setup

docker run -it -v mira-data:/opt/vault -v mira-userdata:/opt/mira/app/data -v mira-pgdata:/var/lib/postgresql/17/main -v mira-valkey:/var/lib/valkey -v mira-backups:/opt/mira/backups -p 1993:1993 mira:latest

Headless (non-interactive)

docker run -e MIRA_PROVIDER_KEY=lr_xxx -v mira-data:/opt/vault -v mira-userdata:/opt/mira/app/data -v mira-pgdata:/var/lib/postgresql/17/main -v mira-valkey:/var/lib/valkey -v mira-backups:/opt/mira/backups -p 1993:1993 mira:latest

Requires building base image first; uses named volumes for persistence

mira-OSS releases & version history

More in AI Chat

Rate this project

Click to rate — no account needed.

0 comments

Posting as guest
Loading comments…