opencode-alchemyst-context

opencode plugin for AlchemystAI Context Layer — gives opencode agents persistent, retrievable memory backed by Alchemyst.

Install

# via npm
npm install opencode-alchemyst-context

# via GitHub
npm install https://github.com/Alchemyst-ai/opencode-alchemyst-context.git

Add to opencode.json:

{
  "plugin": ["opencode-alchemyst-context"]
}

Setup

On first use, tell the agent to configure the API key:

alchemyst_configure: set my key to <your-jwt-bearer-token>

This saves your configuration to ~/.config/opencode-alchemyst/config.json. It persists across sessions — you only need to do this once.

You can also pre-configure by creating the file manually:

{
  "apiKey": "your-jwt-bearer-token",
  "groupName": ["opencode", "my-project"]
}

The config file stores all connection settings (not just the key). Environment variables take priority over saved values — useful for per-session overrides without editing the file.

Env var	Required	Default	Priority
ALCHEMYST_BASE_URL	no	https://platform-backend.getalchemystai.com	overrides config.json
ALCHEMYST_DEFAULT_SCOPE	no	internal	overrides config.json
ALCHEMYST_GROUP_NAME	no	opencode (first element)	overrides config.json

Tools

alchemyst_configure

Set the Alchemyst API key and optionally override connection settings. Required before using any other tool.

apiKey: "your-jwt-bearer-token"
baseUrl: "https://staging.example.com"       # optional
defaultScope: "external"                      # optional
groupName: ["my-team", "project-x"]           # optional, string[]

All values are saved to ~/.config/opencode-alchemyst/config.json. Prompts you for permission before saving. Environment variables for the same settings will override saved values.

alchemyst_context_search

Search stored context for relevant documents.

query: "what did we decide about authentication?"

Returns top results with relevance scores. Use when you need to check if decisions or reference material already exist before answering.

alchemyst_context_add

Save a decision, snippet, or convention to context.

content: "We decided to use Auth0 for all customer-facing auth."
context_type: "instruction"

Runs content through a security guard (blocking keys, tokens, passwords) before sending. Returns the new context ID on success.

alchemyst_memory_add

Persist conversation turns keyed by session ID.

contents: [{ content: "User asked about deployment pipeline..." }]

Session ID is pulled automatically from the opencode session — the model cannot spoof or collide session IDs.

alchemyst_memory_update

Update previously saved memory entries for the current session.

alchemyst_context_ask

Ask a grounded question and get a synthesized answer.

query: "What's our policy on dependency versions?"
steeringPrompt: "answer in one sentence"

Security

The plugin includes a heuristic outbound content guard that checks for:

• Private key headers (BEGIN RSA PRIVATE KEY, etc.)
• AWS access keys (AKIA...)
• OpenAI-style API keys (sk-...)
• GitHub tokens (ghp_..., gho_..., ghu_...)
• Generic api_key assignments
• SECRET/TOKEN/PASSWORD environment assignments

This is a blunt heuristic, not a compliance control. Matched content is blocked before any network request, and the tool returns a clear message about which category was detected. No matched text is echoed back.

Credential handling

The API key is stored in ~/.config/opencode-alchemyst/config.json with restricted permissions (0600). It is never logged, included in error messages, or echoed to the model. Use the alchemyst_configure tool to set or update the key at runtime — no environment variable needed.

If no key is configured, the plugin still loads but every tool returns a clear error message instead of crashing the session.

Development

# Install deps
npm install

# Typecheck
npm run typecheck

# Test
npm test

# Build
npm run build

License

MIT