@evalstate
huggingface/hf-mcp-server
huggingface/upskill
huggingface/skills
fast-agent
Models are placed in an environment, given a task and scored with a reward function:
mini-SWE-Agent: A single 100 line python and single freeform (non JSON) tool can score 76.0% on SWE-Bench!
It's hard to compete against that efficiency.
General Purpose Agent Harnesses are given direct Shell access
Fewer pre/post Tool/LLM Stop checks/hacks to keep model on-track.
Snapshot/Checkpointing techniques (AgentFS, Execution Monitoring)
Remote runtime environments (e.g. Codex Web, Claude Code)
Simple navigable, native hierarchy of content
Reusable procedures become strong scaffolding for capable models
Bash is token dense and unsurprising compared to custom JSON Tools / mid-context tool enablement
Between deterministic program and documentation.
Once models can discover, recover, and keep going, a “skill” becomes a practical acceleration layer rather than a brittle scripted hack.
Dynamic Space Tool: 45 tokens
MCP provides an inference gateway to thousands of specialized and custom models covering Audio, Video, Text, 3D Models, Environments and more.
MCP provides Authentication and Multimodal support.
Qwen 3.5-35B-A3B Flux.1-Krea-Dev Qwen-Edit-2509-Multiple-angles-LoRA Wan2.2 First/Last Frame
Qwen 3.5-35B-A3B
Flux.1-Krea-Dev
Qwen-Edit-2509-Multiple-angles-LoRA
Wan2.2 First/Last Frame
A model with access to general purposes tools has crossed into a very real form of code mode.
Bash provides a general purpose, token dense-execution language.
Task-specific tools generated on demand. Example: HF Tool Builder navigates OpenAPI spec to build composable CLI tools.
Some models are trained to use code tools natively, and are bundled with interpreters.
A common pattern:
Options from YOLO, Local/Remote containers, exe.dev-style or lightweight sandboxes (Monty, Just-Bash). Simple persistent storage (e.g. HF Buckets)
Mixed Model workloads handle different modalitites, specializations and price points. Token efficient task agent delegation.
Increasingly absorb search, tools, code, and state into one bundled execution surface.
Client provided tools, enables "follow along" in editors
Listing, Resumption and Rehydration of Agent sessions
Agent Results and Tool Status stream, are cancellable
Uses MCP Data Model. Client sends MCP Sever Configurations
It defines a shared schema, and tooling layer that enable a unified experience for calling language models, streaming results, and composing agentic workflows—independent of provider.
shell
local_shell
code_interpreter
apply_patch
web_search
etc..
Supports Consumer, Enterprise and Developer use-cases.
Single URL to install authenticated JSON tools across thousands of clients
MCP's "fit" features weren't present at launch!
URI/Resources based extensions deliver innovation and extensibility...
...Which enabled rapid MCP Apps distribution on a solid support base.
Host applications with Shell tool reduce the need for STDIO Servers.
In many cases for local running tools such as Apify mcp-cli or Pete Steinberger's MCPorter offer a better experience for MCP usage.
Distribution via MCPB is one potential advantage
Simple one-shot server design meant that distribution of ideas was more important than code.