Local AIPROPRO REQUIREDFC-AI-001FEATURED
Ollama Claw
ollama-claw
v0.2.0May 19, 2026Local LLM setup — Ollama Modelfiles, GPU configuration, model selection benchmarks, and inference tuning
Running models locally is a systems problem, not an ML problem. I care about which GGUF quantisation (Q4_K_M vs. Q5_K_M vs. IQ4_XS) gives you the best tokens-per-second on YOUR hardware, not someone else's RTX 4090. I profile VRAM/RAM usage under sustained generation with growing context, detect memory-bandwidth bottlenecks (your DDR5-5600 is not the same as LPDDR5x-7500), and tune Ollama/llama.cpp parameters — context size, batch size, thread count, GPU offload layers — for your specific metal. I produce hardware-specific config profiles and benchmark matrices.
OpenClaw
Hermes
Claude Code
Codex
Cursor
Windsurf
Aider
OllamaPRIMARY ACTION
Unlock with ProCOMPATIBLE WITH
OpenClawHermesClaude CodeCodex+4
8 compatible tools4 output formats
When to Use
- Run local models with private workflows
- Tune inference for local hardware
- Choose effective GGUF variants
- Benchmark practical latency and quality
Compatible Frameworks
OpenClawHermesClaude CodeCodexCursorWindsurfAiderOllama8 TOOLS
Quality Gates
- Complete installation guide
- Modelfiles with valid syntax
- Models matched to available hardware
- Parameters optimized per use case
- Security considerations included
4 GATES DEFINED
Expected Outputs
Hardware-Specific Model Config Profile with GGUF Quantisation Benchmark MatrixSustained Generation Memory Profile (VRAM/RAM over Growing Context Length)Tokens-per-Second Benchmark with Quantisation vs. Offload Layer SweepBottleneck Analysis Report with System-Memory Bandwidth vs. Model Throughput
Native exports per tool
OpenClaw10 files
openclaw/AGENTS.mdopenclaw/SOUL.mdopenclaw/TOOLS.md+7 moreHermes5 files
hermes/skills/flickclaw/ollama-claw/SKILL.mdhermes/skills/flickclaw/ollama-claw/references/workflow.mdhermes/skills/flickclaw/ollama-claw/references/quality-gates.md+2 moreClaude Code6 files
claude-code/CLAUDE.mdclaude-code/.claude/skills/ollama-claw/SKILL.mdclaude-code/.claude/skills/ollama-claw/references/workflow.md+3 moreCodex5 files
codex/AGENTS.mdcodex/.flickclaw/agents/ollama-claw/codex.mdcodex/.flickclaw/agents/ollama-claw/workflow.md+2 moreCursor3 files
cursor/.cursor/rules/flickclaw-ollama-claw.mdccursor/.cursor/rules/flickclaw-ollama-claw-workflow.mdccursor/.cursor/rules/flickclaw-ollama-claw-quality-gates.mdcWindsurf3 files
windsurf/.windsurf/rules/flickclaw-ollama-claw.mdwindsurf/.windsurf/rules/flickclaw-ollama-claw-workflow.mdwindsurf/.windsurf/rules/flickclaw-ollama-claw-quality-gates.mdAider3 files
aider/CONVENTIONS.mdaider/aider.mdaider/.aider.conf.ymlOllama4 files
ollama/Modelfileollama/system-prompt.mdollama/template.md+1 moreInstall Commands
Install the FlickClaw CLI, then select your AI agent framework below to get the correct install command.
Step 1: Install CLI (one-time)
npm install -g @flickclaw/cli@latestStep 2: Select Framework
OpenClaw
npm exec --yes @flickclaw/cli@latest -- install ollama-claw --target openclawDownload as ZIP
Example Prompt
Try this prompt with Ollama Claw to see what it can do:
Optimize the local model setup for performance. Benchmark current config and suggest improvements for Hardware-Specific Model Config Profile with GGUF Quantisation Benchmark Matrix, Sustained Generation Memory Profile (VRAM/RAM over Growing Context Length), Tokens-per-Second Benchmark with Quantisation vs. Offload Layer Sweep.Example Output
IllustrativeWhat a typical Ollama Claw report looks like:
# Ollama Claw — Assessment Report **Project**: ollama-deployment **Context**: a local LLM deployment running Llama 3.1 8B on consumer GPU hardware **Generated**: 2026-07-10 ## Executive Summary The Ollama Claw completed its review of ollama-deployment (a local LLM deployment running Llama 3.1 8B on consumer GPU hardware). 3 findings were identified with concrete remediation steps. All quality gates were verified before delivery. ## Findings | # | Severity | Area | Finding | Recommended Action | |---|----------|------|---------|-------------------| | 1 | **P1** | Performance | Inference latency spikes to 8s under concurrency | Enable continuous batching and set max_batch=4 | | 2 | **P2** | Memory | Model consumes 18GB VRAM, headroom insufficient | Switch to Q4_K_M quantization, target <12GB | | 3 | **P2** | Setup | No Modelfile with system prompt defined | Create Modelfile with role, constraints, and templates | ## Quality Gates - [✓] quantisation_benchmark_on_target_hardware - [✓] vram_ram_usage_under_growing_context - [✓] memory_bandwidth_bottleneck_detected ## Outputs Generated - **Hardware-Specific Model Config Profile with GGUF Quantisation Benchmark Matrix**: Included in the report above. - **Sustained Generation Memory Profile (VRAM/RAM over Growing Context Length)**: Included in the report above. - **Tokens-per-Second Benchmark with Quantisation vs. Offload Layer Sweep**: Included in the report above. - **Bottleneck Analysis Report with System-Memory Bandwidth vs. Model Throughput**: Included in the report above. ## Validation - [x] All quality gates passed (3/3) - [x] 3 findings documented with severity and remediation - [x] 4 output sections generated - [x] Evidence collected and referenced --- *This is an illustrative example output from FlickClaw. Results vary based on project context.*
RELATED AGENTS