A recursive-language-model investigation agent with a desktop GUI and terminal interface. OpenPlanter ingests heterogeneous datasets — corporate registries, campaign finance records, lobbying disclosures, government contracts, and more — resolves entities across them, and surfaces non-obvious connections through evidence-backed analysis. It operates autonomously with file I/O, shell execution, web search, and recursive sub-agent delegation.

Pre-built binaries are available on the Releases page:

• macOS — .dmg
• Windows — .msi
• Linux — .AppImage

The desktop app (openplanter-desktop/) is a Tauri 2 application with a three-pane layout:

• Sidebar — Session management, provider/model settings, and API credential status
• Chat pane — Conversational interface showing the agent's objectives, reasoning steps, tool calls, and findings with syntax-highlighted code blocks
• Knowledge graph — Interactive Cytoscape.js visualization of entities and relationships discovered during investigation. Nodes are color-coded by category (corporate, campaign-finance, lobbying, contracts, sanctions, etc.). Click a source node to open a slide-out drawer with the full rendered wiki document.

• Live knowledge graph — Entities and connections render in real time as the agent works. Switch between force-directed, hierarchical, and circular layouts. Search and filter by category.
• Wiki source drawer — Click any source node to read the full markdown document in a slide-out panel. Internal wiki links navigate between documents and focus the corresponding graph node.
• Session persistence — Investigations are saved automatically. Resume previous sessions or start new ones from the sidebar.
• Background wiki curator — A lightweight agent runs in the background to keep wiki documents consistent and cross-linked.
• Multi-provider support — Switch between OpenAI, Anthropic, OpenRouter, Cerebras, and Ollama (local) from the sidebar.

cd openplanter-desktop

# Install frontend dependencies
cd frontend && npm install && cd ..

# Run in development mode
cargo tauri dev

# Build distributable binary
cargo tauri build

Requires: Rust stable, Node.js 20+, and platform-specific Tauri dependencies (see Tauri prerequisites).

The Python CLI agent can be used independently of the desktop app.

# Install
pip install -e .

# Configure API keys (interactive prompt)
openplanter-agent --configure-keys

# Launch the TUI
openplanter-agent --workspace /path/to/your/project

Or run a single task headlessly:

openplanter-agent --task "Cross-reference vendor payments against lobbying disclosures and flag overlaps" --workspace ./data

# Add your API keys to .env, then:
docker compose up

The container mounts ./workspace as the agent's working directory.

Ollama runs models locally with no API key. Install Ollama, pull a model (ollama pull llama3.2), then:

openplanter-agent --provider ollama
openplanter-agent --provider ollama --model mistral
openplanter-agent --provider ollama --list-models

The base URL defaults to http://localhost:11434/v1 and can be overridden with OPENPLANTER_OLLAMA_BASE_URL or --base-url. The first request may be slow while Ollama loads the model into memory; a 120-second first-byte timeout is used automatically.

Additional service keys: EXA_API_KEY (web search), VOYAGE_API_KEY (embeddings).

All keys can also be set with an OPENPLANTER_ prefix (e.g. OPENPLANTER_OPENAI_API_KEY), via .env files in the workspace, or via CLI flags.

The agent has access to 19 tools, organized around its investigation workflow:

Dataset ingestion & workspace — list_files, search_files, repo_map, read_file, write_file, edit_file, hashline_edit, apply_patch — load, inspect, and transform source datasets; write structured findings.

Shell execution — run_shell, run_shell_bg, check_shell_bg, kill_shell_bg — run analysis scripts, data pipelines, and validation checks.

Web — web_search (Exa), fetch_url — pull public records, verify entities, and retrieve supplementary data.

Planning & delegation — think, subtask, execute, list_artifacts, read_artifact — decompose investigations into focused sub-tasks, each with acceptance criteria and independent verification.

In recursive mode (the default), the agent spawns sub-agents via subtask and execute to parallelize entity resolution, cross-dataset linking, and evidence-chain construction across large investigations.

openplanter-agent [options]

Use --default-model, --default-reasoning-effort, or per-provider variants like --default-model-openai to save workspace defaults to .openplanter/settings.json. View them with --show-settings.

Keys are resolved in this priority order (highest wins):

1. CLI flags (--openai-api-key, etc.)
2. Environment variables (OPENAI_API_KEY or OPENPLANTER_OPENAI_API_KEY)
3. .env file in the workspace
4. Workspace credential store (.openplanter/credentials.json)
5. User credential store (~/.openplanter/credentials.json)

All runtime settings can also be set via OPENPLANTER_* environment variables (e.g. OPENPLANTER_MAX_DEPTH=8).

openplanter-desktop/         Tauri 2 desktop application
  crates/
    op-tauri/                 Tauri backend (Rust)
      src/commands/           IPC command handlers (agent, wiki, config)
    op-core/                  Shared core library
  frontend/                   TypeScript/Vite frontend
    src/components/           UI components (ChatPane, GraphPane, InputBar, Sidebar)
    src/graph/                Cytoscape.js graph rendering
    src/api/                  Tauri IPC wrappers
    e2e/                      Playwright E2E tests

agent/                        Python CLI agent
  __main__.py                 CLI entry point and REPL
  engine.py                   Recursive language model engine
  runtime.py                  Session persistence and lifecycle
  model.py                    Provider-agnostic LLM abstraction
  builder.py                  Engine/model factory
  tools.py                    Workspace tool implementations
  tool_defs.py                Tool JSON schemas
  prompts.py                  System prompt construction
  config.py                   Configuration dataclass
  credentials.py              Credential management
  tui.py                      Rich terminal UI
  demo.py                     Demo mode (output censoring)
  patching.py                 File patching utilities
  settings.py                 Persistent settings

tests/                        Unit and integration tests

cd openplanter-desktop

# Development mode (hot-reload)
cargo tauri dev

# Frontend tests
cd frontend && npm test

# E2E tests (Playwright)
cd frontend && npm run test:e2e

# Backend tests
cargo test

# Install in editable mode
pip install -e .

# Run tests
python -m pytest tests/

# Skip live API tests
python -m pytest tests/ --ignore=tests/test_live_models.py --ignore=tests/test_integration_live.py

Requires Python 3.10+. Dependencies: rich, prompt_toolkit, pyfiglet.

MIT — see LICENSE for details.