Memex Journal
Turn fragmented text, photos, and voice notes into local timeline cards and knowledge insights.
- Source repo
- memex-lab/memex
- Stars
- ★ 827
- Last updated
- 2d ago
- License
- GPL-3.0
- Primary language
- Dart
- FA score
- 50/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformOpenAI API · Claude APIChatGPT (Partial support)
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A person tracking tasks, routines, events, and goals in scattered notes can have those fragments organized into task, event, routine, and progress cards.
- Not a fit if
- Users who need a desktop or web app; this is iOS and Android only
- Teams unwilling to supply their own API keys and pay model usage
- Source review
- 50/100 · Major gaps
What does this agent do, and when should you use it?
Memex is an open-source, local-first AI journal for iOS and Android that is designed for capturing fragments rather than composing polished daily entries. The Flutter application uses Drift (SQLite), the local filesystem, and a Provider + MVVM presentation architecture. Its Memex Agent coordinates specialist skills and child agents to turn text, images, voice, and shared files into timeline cards, a P.A.R.A. Markdown knowledge base, schedule state, and insight cards. Comment and companion agents add character comments and one-to-one chat, supported by long-term memory services. A Custom Agent System also lets users configure event triggers, SKILL.md-based behavior, workspaces, models, and synchronous or queued execution.
Memex accepts text, images, voice, and shared files, then prepares assets through transcription, document text extraction, image encoding, and metadata handling before Memex Agent orchestration. It invokes specialist skills or child agents such as timeline_card, pkm, schedule, and knowledge_insight: timeline_card produces Cards YAML and UI config, pkm creates a P.A.R.A. Markdown knowledge base, schedule updates schedule state, and knowledge_insight creates cross-record insight cards. Background work can include character comments, long-term memory curation, and event-driven custom agents. Custom agents can run after a timeline card is saved, a card comment is posted, card UI configuration changes, or local data changes; they support dependsOn chains, retries, scoped working directories, and synchronous or asynchronous execution. Skills can execute JavaScript, including fetch() calls to external APIs. AI features are configured through listed services including Gemini, OpenAI, Claude, Bedrock, Memex AI, OpenRouter, and Ollama.
- A person tracking tasks, routines, events, and goals in scattered notes can have those fragments organized into task, event, routine, and progress cards.
- A researcher or knowledge worker collecting articles, snippets, quotes, links, and conversations can use the pkm skill to file material into a P.A.R.A. Markdown knowledge base.
- Someone reviewing a long record of life activity can use chart, narrative, map, route, and timeline insight cards to identify patterns and reconstruct context.
- A privacy-conscious journaler who wants device-resident records can use the local filesystem and SQLite storage while selecting their own model service.
- An advanced mobile user building personal automations can create custom agents with local event triggers, SKILL.md instructions, individual workspaces, and dependency ordering.
How do you install or deploy this agent?
Install from the App Store, Google Play, or the Android APK published through GitHub Releases. Source builds require Flutter SDK ≥ 3.6.0:
git clone https://github.com/memex-lab/memex.git
cd memex
flutter pub getFor iOS, also run:
cd ios && pod install && cd ..
Then launch a development build with:
flutter run --flavor globalDevFor Android local development, use globalDev or cnDev to keep package IDs and app data isolated.
How do you use this agent?
At first launch, open the avatar icon → Model Configuration. Choose a provider such as Gemini, OpenAI, Claude, DeepSeek, Memex AI, or Ollama, then supply the API key and base URL, sign in through OAuth, or connect Memex AI as required by that provider. Models can be configured separately for each agent. Capture text, photos, or voice to begin creating organized records. To attach location context, open Personal center → Settings → Location, enable “Attach current location to chat,” then choose OpenStreetMap / Nominatim or enter an Amap API Key.
What are this agent's strengths and limitations?
- Records, cards, and knowledge are stored in the device filesystem and SQLite, with standard Markdown export for moving the archive elsewhere.
- The documented processing path has distinct outputs: Cards YAML and UI config, P.A.R.A. Markdown, schedule state, and cross-record insight cards.
- It supports multiple model services, including OpenAI, Claude, Gemini, Bedrock, OpenRouter, and local Ollama, with per-agent model selection.
- Its Custom Agent System supports local event triggers, dependsOn ordering, sync or async execution, automatic retries, scoped workspaces, and JavaScript skills.
- AI features require a configured model service, which may require an API key, base URL, OAuth sign-in, or a Memex AI account.
- Journal data remains local, but prompts are sent directly to the selected model provider; the optional Memex AI service proxies model requests.
- The ChatGPT and Google OAuth options are documented as unofficial and at the user's own risk.
- Source builds are explicitly documented only for iOS and Android and require Flutter; iOS additionally requires Xcode and pod install, while Android requires Android Studio.
How does this agent compare with similar options?
Unlike traditional journaling apps that ask for polished daily entries, Memex is organized around fragments such as text, photos, and voice that are processed by a multi-agent workflow. Unlike a general note-taking destination, it is designed to settle organized records into interconnected Markdown files that can be exported.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Memex Journal This agent | 50 · Major gaps | Desktop appFree + model costs | ★ 827 | 2d ago | Dart | OpenAI API · Claude API |
| Kanzo Board | 49 · Major gaps | CLIFree | ★ 1.1k | 1d ago | Rust | Codex · Claude Code |
| Field Theory CLI | 50 · Major gaps | CLIFree + model costs | ★ 2.1k | 2mo ago | TypeScript | Codex · Claude Code |
| Lotti Private Logbook | 80 · Good | Desktop appFree + model costs | ★ 1.2k | today | Dart | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: README clearly states local storage, no server storing journal, optional proxy does not store data; provides app lock and biometric auth; supports custom working directories and skill execution scopes. However, lacks implementation details for least privilege (e.g., Android/iOS permission manifests not provided); user confirmation mechanism not explicit (e.g., whether AI actions require confirmation); data flow transparency described but not code-verified; sensitive data handling mentioned but not deep; dependency security not mentioned (no vulnerability scanning or version pinning); external effects (e.g., network requests) not explicitly restricted; rollback mechanism not mentioned; source attribution not clear. Thus most criteria score low.
Evidence shows: README and project structure are consistent, described features match architecture; but dependency availability not verified (e.g., third-party LLM providers' availability not guaranteed); failure messages not provided with examples. Thus self-consistency scores higher, others lower.
Evidence shows: clear target users (iOS/Android users) and use cases (journaling, AI organization); capability boundaries described (e.g., custom agents, skills, working directories); trigger conditions enumerated (event-driven); environment fit described (multi-platform, multi-LLM providers). But lacks specific configuration examples and boundary condition tests. Thus scores moderate.
Evidence shows: README structure clear, includes installation, configuration, development, architecture; install notes detailed; naming stable (e.g., Memex, Agent); but lacks FAQ and examples; known limitations not explicitly listed; license is GPL-3.0, but no changelog provided; maintenance responsibility not clear. Thus scores moderate.
Evidence shows: output is Markdown files, easy to use; marginal value high (AI organization, knowledge base, companion); cost-benefit reasonable (bring your own LLM, local storage). But no real-world usage cases or performance data. Thus scores moderate.
Evidence shows: README claims (e.g., local-first, multi-LLM support) have partial code support (e.g., project structure), but no test results or independent verification; cross-source corroboration insufficient (only README and few tests); facts and inferences not clearly separated. Thus scores low.
- Publisher identity unverified; be cautious about supply chain risks.
- Dependency security not mentioned; recommend checking for vulnerabilities and pinning versions.
- User confirmation mechanism unclear; AI actions may execute automatically, raising privacy and misoperation risks.
- External network requests (e.g., fetch in skills) may pose data exfiltration risks; review permissions.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Android Early build from main. - Commit: 5ecbe4c59aaac1f1d62a5f82b0cd495a2b4c1fa2 - Date: 2026-11-07 Asia/Shanghai - Channel: Early - Version: 1.0.39-early.20261107.5ecbe4c+1793990712 - Packages: - Global: com.memexlab.memex.early - CN: com.memexlab.memex.cn.early Changes since previous Early prerelease Compared with android-early-20261106-5ecbe4c. Full Changelog: https://github.com/memex-lab/memex/compare/android-early-20261106-5ecbe4c...android-early-20261107-5ecbe4c Changes since latest Release Compared with v1.0.37. What's Changed …View on GitHub ↗ -
v1.0.37 Memex v1.0.37
-
v1.0.36 Memex v1.0.36