Ollama Draft

Ollama · reviewed 22 November 2026 · its card

Ollama lets people run open models on their own computer or in Ollama's cloud, and point desktop apps, coding agents and their own applications at them: "Use open models in your desktop apps and coding agents, or build them into your application" (S1). The repository describes it as "Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models" (S9). It runs on macOS, Windows and Linux as an app, a CLI and a local API, with compatible OpenAI and Anthropic APIs (S2, S12, S1). The software is open source under the MIT licence (S9); cloud models are sold on Free, Pro ($20 a month), Max ($100 a month) and an early-access Team plan from $500 a month (S11). The latest release seen is v0.40.17, of 10 November 2026 (S10). It is a model runtime, the neighbour of model rather than an agent harness: the pages read describe serving models and launching other agents, not running an agent loop of its own.

How to read a product card.

Capabilities

A product page in four labelled parts: what it is and who makes it; the capabilities table, with columns term, value, detail and sources, and the values yes, partial, no and unknown shown by the matrix's glyphs; the state of play, dated, newest first; and the sources, each with the date it was read.
How to read a product page. Every product has the same parts in the same order.
CapabilityValueDetailSources
Model Yes Runs models locally or "in Ollama's cloud"; "Ollama supports tool calling (also known as function calling)" in its API; MLX by default on Apple silicon S2 S6 S7 S10
Session Unknown Not found; ollama run chats with a model, with no saved agent sessions described S5
Subagent Unknown Not found None
Agent team Unknown Not found None
Skill Unknown Not found None
Instruction file Unknown Not found None
Plugin Unknown Not found None
MCP Unknown Not found as a client; it offers a web-search MCP server "in any MCP client through the Python MCP server" S8
Execution environment Unknown Not found None
Client Unknown The app and ollama run chat with a model; no agent sessions described S2 S5
Agent host Partial ollama launch "Configure[s] and launch[es] external applications to use Ollama models": Claude Code, Codex, OpenCode, Droid, Kimi Code CLI, OpenClaw, Hermes Agent and others, offering to install some; nothing shows it owning their sessions S2 S3 S5 S10
Channel Unknown Not found; its Claude Code page passes Claude Code's own --channels flag through ollama launch S4
Cross-session messaging Unknown Not found None
ACP Unknown Not found None
AHP Unknown Not found None
A2A Unknown Not found None
Bot Unknown Not found; OpenClaw and Hermes Agent, launched through it, are other makers' assistants S3

State of play

How things stand, newest first, as the card records it.

  • Version 0.40.17, the latest release seen: "ollama launch claude supports Claude Code's deferred tool loading". S10
  • Version 0.40.16: "ollama launch claude always routes Claude Code to the selected model". S10
  • Version 0.40.15: "ollama launch now supports Kimi Code CLI", and running ollama with no arguments opens the welcome launcher. S10
  • Version 0.40.0: "Models run on MLX on Apple Silicon by default". S10

Sources

IdSourceRead
S1 Ollama, Ollama documentation
S2 Ollama, Quickstart
S3 Ollama, Integrations overview
S4 Ollama, Claude Code integration
S5 Ollama, CLI reference
S6 Ollama, Cloud
S7 Ollama, Tool calling
S8 Ollama, Web search
S9 Ollama, ollama/ollama repository
S10 Ollama, ollama/ollama releases
S11 Ollama, Pricing
S12 Ollama, Download