{
  "title": "Local models",
  "description": "Run against a model hosted on the operator's machine.",
  "slug": "local-models",
  "locale": "en",
  "seoTitle": "Local models — Can My Agent Use",
  "socialTitle": "Local models",
  "socialDescription": "Compare connections to locally hosted models.",
  "llmSummary": "Local model support lets a product use a model hosted on the operator's machine. Model servers, formats, authentication, and feature limitations are qualifiers.",
  "audience": "Engineers comparing chat, desktop, and CLI agent harnesses.",
  "contentKind": "feature",
  "status": "published",
  "tags": [
    "runtime"
  ],
  "updated": "2026-08-28T00:00:00.000Z",
  "published": "2026-08-28T00:00:00.000Z",
  "category": "models-context",
  "summary": "Run against a model hosted on the operator's machine.",
  "specLabel": "Common product term",
  "aliases": [
    "self-hosted model",
    "local inference",
    "on-device model"
  ],
  "capabilityKind": "atomic",
  "parent": "models-and-context",
  "related": [
    "model-selection"
  ],
  "relations": [],
  "highlight": false,
  "notes": [
    {
      "id": 1,
      "text": "Product cells without reviewed public evidence remain unknown; a configurable remote API endpoint alone is not treated as proof of on-device inference."
    },
    {
      "id": 2,
      "text": "Evidence checked 2026-08-28: VS Code documents BYOK support for fully offline local models such as Ollama and an experimental setting that enables BYOK models in Agent Host sessions."
    },
    {
      "id": 3,
      "text": "Evidence checked 2026-08-28: Continue documents local Ollama configuration, automatic discovery of installed models, offline operation, and model-role limitations."
    },
    {
      "id": 4,
      "text": "Evidence checked 2026-08-28: aider documents connecting directly to a local Ollama endpoint and selecting local models from the CLI."
    },
    {
      "id": 5,
      "text": "Evidence checked 2026-08-28: Cline documents fully offline local-model operation with either Ollama or LM Studio and calls out hardware and tool-use constraints."
    },
    {
      "id": 6,
      "text": "Evidence checked 2026-08-28: OpenCode documents local Ollama discovery and explicit configuration for Ollama, vLLM, and other locally hosted OpenAI-compatible servers."
    }
  ],
  "issues": [],
  "resources": [
    {
      "title": "Methodology",
      "href": "/methodology",
      "kind": "note"
    },
    {
      "title": "Ollama — Quickstart",
      "href": "https://docs.ollama.com/quickstart",
      "kind": "docs",
      "publisher": "Ollama",
      "reviewedAt": "2026-08-28"
    },
    {
      "id": "microsoft-vscode-local-language-models",
      "title": "Microsoft — AI language models in VS Code",
      "href": "https://code.visualstudio.com/docs/agent-customization/language-models",
      "kind": "docs",
      "publisher": "Microsoft",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Bring your own language model key and local Ollama models"
    },
    {
      "id": "continue-local-ollama-models",
      "title": "Continue — Using Ollama with Continue",
      "href": "https://docs.continue.dev/guides/ollama-guide",
      "kind": "docs",
      "publisher": "Continue",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Configure Ollama with Continue and model capabilities"
    },
    {
      "id": "aider-local-ollama-models",
      "title": "aider — Ollama",
      "href": "https://aider.chat/docs/llms/ollama.html",
      "kind": "docs",
      "publisher": "aider",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Local Ollama endpoint and model selection"
    },
    {
      "id": "cline-local-models",
      "title": "Cline — Local models overview",
      "href": "https://docs.cline.bot/running-models-locally/overview",
      "kind": "docs",
      "publisher": "Cline",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Running models locally with Cline"
    },
    {
      "id": "opencode-local-models",
      "title": "OpenCode — Models",
      "href": "https://opencode.ai/v2/docs/models",
      "kind": "docs",
      "publisher": "OpenCode",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Local models"
    }
  ],
  "support": [
    {
      "harness": "vscode-copilot",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            2
          ],
          "target": {
            "kind": "dated-documentation",
            "revision": "current VS Code language-model documentation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "preview-enabled",
          "qualifiers": [
            {
              "type": "experimental",
              "value": "using BYOK models in Agent Host sessions requires chat.agentHost.byokModels.enabled and an Agent Host restart"
            },
            {
              "type": "runtime",
              "value": "local Ollama use now requires the official Ollama extension; some VS Code features such as semantic search and embeddings can still require a GitHub account or external service"
            }
          ],
          "evidence": [
            {
              "resourceId": "microsoft-vscode-local-language-models",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ],
          "stage": "experimental"
        }
      ]
    },
    {
      "harness": "continue",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            3
          ],
          "target": {
            "kind": "dated-documentation",
            "revision": "current Continue Ollama guide",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "local-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "local model quality and advertised tool calling determine whether Agent mode, edit, apply, embeddings, and other roles work reliably"
            }
          ],
          "evidence": [
            {
              "resourceId": "continue-local-ollama-models",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "aider",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            4
          ],
          "target": {
            "kind": "dated-documentation",
            "revision": "current aider Ollama documentation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "local-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "weaker local models may fail aider's code-edit protocol; the documented default endpoint is the operator's local Ollama server"
            }
          ],
          "evidence": [
            {
              "resourceId": "aider-local-ollama-models",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "cline",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            5
          ],
          "target": {
            "kind": "dated-documentation",
            "revision": "current Cline local-model guide",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "air-gapped",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "requires a locally running Ollama or LM Studio endpoint and sufficient hardware; feature completeness depends on the model's tool-use behavior and context capacity"
            }
          ],
          "evidence": [
            {
              "resourceId": "cline-local-models",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "opencode",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            6
          ],
          "target": {
            "kind": "dated-documentation",
            "revision": "current OpenCode local-model documentation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "local-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "OpenCode can discover local Ollama models and configure local Ollama, vLLM, or OpenAI-compatible servers; declared capabilities and limits must match the actual model"
            }
          ],
          "evidence": [
            {
              "resourceId": "opencode-local-models",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    }
  ],
  "html": "/features/local-models",
  "markdown": "/features/local-models.md",
  "json": "/api/v1/features/local-models.json",
  "body": "This row asks whether the exact harness can send model requests to an operator-controlled local inference runtime. A cloud model with a desktop client is not a local model, and a configurable remote API base URL does not prove on-device execution.\n\nEvidence should identify the supported runtimes or protocols, model-discovery behavior, tool-calling and multimodal constraints, authentication, and whether any prompts, telemetry, or embeddings still leave the device.\n\nThe matrix below lists published **web**, **desktop**, and **CLI** surfaces. Unreviewed cells remain **unknown** until a dated note and public source support a more specific status.\n\nUse the Markdown and JSON twins if you are an agent reading this site."
}