# renezander030/agentic-task-system [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/renezander030/agentic-task-system  
**GitHub Stars:** 12  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/renezander030-agentic-task-system

## Description
Turns your task manager into agent memory: hybrid (dense + sparse + keyword, RRF) retrieval over TickTick or an Obsidian vault via a six-method adapter contract. MCP server + CLI, no vector DB to build or maintain. npm i -g @reneza/ats-cli

## Claude Desktop Quick Installation
Heuristic fallback — verify the package name and runner against the repository README before running it. Uses `npx` (confidence: low):

```json
"mcpServers": {
  "agentic-task-system": {
    "command": "npx",
    "args": ["-y","renezander030-agentic-task-system"]
  }
}
```

## Documentation

## What renezander030/agentic-task-system MCP server does

The renezander030/agentic-task-system MCP server, exposed through the `ats` CLI and MCP interface, gives agents a common way to search task records, notes, decisions, and runbooks stored in existing systems. Supported adapter examples include TickTick, Obsidian, Taskmaster, Beads, Notion, GitHub, Airtable, Google, and OKF files. A composite adapter can combine several sources while leaving each source authoritative.

The project has two related layers. The task layer maps backend records into a shared contract for retrieval and lifecycle operations. The knowledge-graph layer, accessed with `ats kg`, stores subject–predicate–object facts with provenance and temporal validity. Agents can propose facts, while human ratification is part of the described workflow. The embedded graph store does not require a separate graph database.

## How it works

Search can use keyword matching, a source adapter's native search, and dense retrieval when an adapter provides it. Available branches are fused with reciprocal rank fusion, and results retain provenance so an operator can identify their origin. The `find --explain` option exposes RRF contributions, while confidence reflects agreement among retrieval branches; `--min-sources N` can require a matching number of sources.

Derived retrieval state includes a five-minute corpus cache by default and, where configured, vector indexes. The TickTick reference adapter can use Qdrant and Ollama for dense retrieval. Without vectors, keyword and native-search branches still operate. Cached records, query text, action before-images, and vector payloads may contain task content, and the project does not apply application-level encryption or runtime redaction to those copies.

The renezander030/agentic-task-system MCP server supports local stdio operation, which does not expose an MCP network port. A hosted deployment can add a bearer-token gateway and private Qdrant or Ollama services, but hosting, persistence, backups, token rotation, and availability remain operator responsibilities. Composite adapter children run inside the same ATS process; delegation separates routing and authentication ownership, not process boundaries.

## Setup and configuration

The README shows a global npm installation of the CLI together with the TickTick adapter:

```bash
npm install -g @reneza/ats-cli @reneza/ats-adapter-ticktick
ats config use ticktick
ats auth login
ats find "deployment runbook"
```

The example uses TickTick, but the project is designed around adapters and also documents composite use across multiple sources. Adapter-specific authentication, field mapping, and deep links remain at the adapter boundary. The provided material does not specify a universal environment-variable configuration or a standalone command for launching the MCP process, so deployments should follow the relevant adapter and project documentation.

## Tools and capabilities

The renezander030/agentic-task-system MCP server and CLI provide these documented capabilities:

- Search across available task and knowledge sources with `find`.
- Combine keyword, native, and optional dense retrieval with RRF.
- Restrict searches by project and inspect ranking details with `--project` and `--explain`.
- Return provenance and degraded-status warnings when retrieval branches or composite corpus operations fail.
- Write through an active adapter when it supports writes, using patch semantics such as append and prepend.
- Preserve write before-images for supported undo workflows and use content hashes for conditional updates.
- Export knowledge-graph data as JSON with provenance or as Cypher for LadybugDB loading.

The task backend remains the canonical record. Caches and vector indexes improve retrieval but are derived copies rather than records users must maintain separately.

_Full upstream README: https://allmcps.com/mcp/renezander030-agentic-task-system/readme_