Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

5.6 KB
Raw Blame

OpenCode Integration

Repowise and OpenCode connect in two independent directions, and it is worth knowing which one you want before reading further. They share nothing but a name.

Direction What it is Where to read
Repowise calls OpenCode The opencode LLM provider. Repowise runs your local OpenCode CLI to generate wiki pages, instead of using an API key. opencode Provider below
OpenCode calls Repowise The opencode agent target. OpenCode gets the repowise MCP tools, so you can ask it about your codebase. OpenCode as an MCP host below

You can use either, both, or neither. Enabling one does not enable the other.

OpenCode as an MCP host

repowise agents add --target=opencode

This registers the repowise MCP server under the mcp key of your opencode.jsonc (or an existing opencode.json) and adds a managed section to AGENTS.md. By default it writes both the repo-local pair at the repo root and the per-machine pair under $XDG_CONFIG_HOME/opencode, falling back to ~/.config/opencode; pass --scope=project or --scope=user for one of them. That config path is the same on Windows, where OpenCode does not read %APPDATA%.

OpenCode is at the Good tier: it gets the MCP tools and the config to reach them, but no hook-level interception and no transcript mining. See the support matrix for what that means in full.

Two things worth knowing:

  • Comments in the config are safe. OpenCode accepts JSONC, and repowise does not rewrite a config it cannot parse as strict JSON. If yours has comments, repowise agents add leaves it alone and tells you to run repowise agents print-config opencode and paste the entry yourself.
  • AGENTS.md is shared with Codex and Hermes. All three agents read the same file, and all three manage the same marker-delimited section of it. Removing one of them leaves the section in place while another is still wired, and says so.

Remove it with repowise agents remove --target=opencode, and check it with repowise agents or repowise doctor. To remove every agent at once, along with the index and anything else repowise has written, see repowise uninstall.

Prerequisites

Install OpenCode:

curl -fsSL https://opencode.ai/install | bash

Then run opencode once to set up your model provider and authentication:

opencode

Verify the CLI is available:

opencode --version

Repowise detects OpenCode automatically: when opencode is in PATH, the interactive provider selection shows it as "available" and it can be used immediately.

opencode Provider

Use opencode when you want Repowise page generation to run through your local OpenCode CLI instead of an API key:

repowise init --provider opencode --yes

You can also persist it:

REPOWISE_PROVIDER=opencode repowise update

The provider runs opencode run --format json with the repository as the subprocess working directory:

cd /absolute/path/to/repo && opencode run --format json

Repowise sends the combined system + user prompt on stdin, parses OpenCode's JSONL output (extracting text from text events and token usage from step_finish events), and treats opencode/* cost as $0.00 because billing is handled by OpenCode's own subscription/auth.

Default model

opencode/default uses OpenCode's configured default model — no --model flag is passed. To use a specific model:

repowise init --provider opencode --model opencode/deepseek-v4-pro

Or use a bare model slug (the opencode/ prefix is optional):

repowise init --provider opencode --model deepseek-v4-pro

Listing available models

opencode models           # all available models
opencode models opencode  # models from the opencode provider

Reasoning

The opencode provider does not pass reasoning effort flags. OpenCode handles reasoning internally through its own model/agent configuration.

Security

The provider enforces several safety measures:

  • Uses asyncio.create_subprocess_exec (no shell), so every argument is passed as a distinct list element — shell injection is impossible.
  • Model names are validated against a safe character set ([a-zA-Z0-9][a-zA-Z0-9._/\-]*), rejecting shell metacharacters before anything reaches the subprocess.
  • All paths are resolved with Path.resolve() before the repo is used as the subprocess working directory.
  • Subprocess execution is serialized via asyncio.Semaphore(1).
  • A 600-second hard timeout with process kill prevents runaway calls.

Comparison with Codex CLI

Aspect opencode codex_cli
CLI command opencode run codex exec
Auth OpenCode providers codex login
Output format JSONL via --format json JSONL via --json
Reasoning modes Not passed (OpenCode manages it) model_reasoning_effort mapping
Sandbox OpenCode manages its own --sandbox read-only
Model discovery opencode models codex debug models --bundled
MCP host integration opencode.jsonc, AGENTS.md (Good tier) .codex/config.toml, hooks, plugin, prompts (Full tier)
API keys stored No No

Official OpenCode Docs

Example walkthrough

Short smoke path: examples/opencode/.