Shrimp Task Manager
Coding · tested 2026-08-26 · re-test due 2026-11-24 · by the Hlido desk, not the vendor
In short: An MCP task manager that gives AI coding assistants long-term memory, structured task decomposition with dependencies, and execution tracking — aimed squarely at the 'assistant forgets across conversations' problem.
5 PASS · 0 FAIL of 5 public-surface claims
Quick answer
Shrimp Task Manager scores 75/100 (STEADY) on Hlido’s independent, hands-on test (reviewed 2026-08-26). STEADY (75) because Shrimp Task Manager maps three real failure modes of AI coding assistants to concrete features — persistent task memory, dependency-aware decomposition, execution tracking, knowledge reuse — delivered Pricing: Open source (free entry point documented).
Shrimp Task Manager names its target precisely: the three ways AI programming assistants fail at multi-step work — memory loss across conversations, structural chaos in complex tasks, and starting every session from scratch — and answers each with a concrete feature. It provides a task-memory function that persists execution history, automatic decomposition of complex tasks into subtasks with explicit dependencies and ordered execution paths, real-time execution-status tracking with progress visualization, and a knowledge-accumulation store that records successful solutions for reuse. Delivered as an MCP server, it plugs into agentic coding assistants rather than being a standalone product, which is the right shape for the job. The documentation is well-structured (pain-points → six core features → workflow → prompt config → installation) and bilingual, and it is open source on GitHub. It lands mid-STEADY because the strengths are described capabilities rather than measured outcomes in this review, there is no independent benchmark of how well decomposition and dependency-tracking actually hold up on real tasks, and no adoption/maintenance signal was captured. As a structured-memory layer for a coding agent, though, it is coherent and well-scoped.
Why STEADY
STEADY (75) because Shrimp Task Manager maps three real failure modes of AI coding assistants to concrete features — persistent task memory, dependency-aware decomposition, execution tracking, knowledge reuse — delivered agent-native over MCP with clear docs. Held mid-STEADY because these are described capabilities, not outcomes measured in this review, with no independent benchmark or captured adoption/maintenance signal.
Public-surface checklist
- PASS Homepage loads (required)
- PASS Primary value prop (required) — Structured task management + long-term memory for AI programming assistants
- PASS Cta present (required) — Get Started / Installation / GitHub
- PASS Pricing or access — Open source on GitHub; MCP install
- PASS Evidence or demo — Pain-points, six-feature breakdown, workflow and prompt-config docs
What we saw
4 screenshots captured by the Hlido engine during the reviewed run (run-a98709bb99713d17-cjo4m06-github-io). Our own captures — not vendor marketing material.
What it does well
- Persistent task memory across conversations — directly targets the assistant's cross-session amnesia
- Automatic decomposition of complex tasks into subtasks with explicit dependencies and ordered execution
- Real-time execution-status tracking with progress visualization and completion reports
- Knowledge-accumulation store that records successful solutions for reuse on similar tasks
- Agent-native as an MCP server; clear, well-structured, bilingual documentation and prompt config
- Open source on GitHub
What it fails at
- Strengths are described features, not outcomes measured in this review
- No independent benchmark of decomposition/dependency-tracking quality on real tasks
- No captured adoption, maintenance-cadence or license detail on the reviewed surface
- Value depends on the host assistant honouring the structured plan it produces
Best for
- Developers whose AI coding assistant loses track of multi-step tasks across a session
- Agentic workflows that need explicit task decomposition, dependencies and progress tracking
- Teams that want a reusable knowledge base of solved tasks feeding future prompts
Not recommended for
- Users wanting a standalone project-management app rather than an MCP layer for an assistant
- Teams that require proven, benchmarked reliability guarantees today
Pricing & access
- ModelOpen source
- Free entry pointYes — a free tier or open-source edition is documented
- Pricing findable on the public surfacePASS Open source on GitHub; MCP install (tested 2026-08-26)
Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-08-26.
Compared to
-
ai-orchestrator
in-assistant-task-memory-vs-cross-agent-orchestration
Both structure multi-step agent work; the orchestrator coordinates agents/flows broadly, while Shrimp is a focused MCP memory-and-decomposition layer for a coding assistant. Orchestrator for cross-agent flow control, Shrimp for in-assistant task memory and breakdown.
-
codebase-memory-mcp
task-state-memory-vs-codebase-memory
codebase-memory-mcp persists knowledge about the codebase; Shrimp persists task state and decomposition. Codebase-memory for what the code is, Shrimp for what the assistant is doing to it.
Agent relevance
MCP Behavioral-testable
Agentic-Commerce Readiness 60/100 · INTEGRABLE
Independent readiness for agent delegation & transaction. How it’s scored · check live
Runs as an MCP server connected to an AI coding assistant; the assistant calls it to plan and decompose tasks, persist task memory across sessions, track execution status, and reference a knowledge base of prior solutions. It is a structured-memory-and-planning layer for the agent rather than a standalone tool.
Agent-friendly score: 8/10
Score over time
The longitudinal record — every point is the score as published on that date. Raw series.
Evidence
- Provides long-term/task memory so assistants remember progress across conversations — source (2026-08-26) verified
- Automatically decomposes complex tasks into subtasks with dependencies and ordered execution — source (2026-08-26) verified
- Six core features incl. execution-status tracking and knowledge accumulation — source (2026-08-26) verified



