# Fiber AI — OpenCode install instructions

You are an AI agent running inside OpenCode. The user asked you to install the Fiber AI plugin. Follow these steps in order. Do not skip steps.

## 1. Confirm the user has a Fiber API key

Ask the user if they already have `FIBER_API_KEY` set in their shell. If they do not:

- Direct them to <https://fiber.ai/app/api> to create one.
- Show them how to persist it:

  ```bash
  echo 'export FIBER_API_KEY=sk_live_your_key_here' >> ~/.zshrc
  source ~/.zshrc
  ```

Do not proceed until the user confirms they have `FIBER_API_KEY` set.

## 2. Register the Fiber MCP servers

OpenCode reads MCP config from one of:

- `./opencode.json` (project-level, preferred for team sharing)
- `~/.config/opencode/opencode.json` (user-global)

Decide with the user which scope they want. Default to project-level.

Add the following entries to the chosen config file's `mcp` object. V3 uses OAuth (browser-based SSO login) instead of an API key - ask the user whether they want to register it; skip it if they only want the API-key path.

```json
{
  "mcp": {
    "fiber-ai-v2": {
      "type": "remote",
      "url": "https://mcp.fiber.ai/mcp/v2",
      "headers": {
        "x-api-key": "{env:FIBER_API_KEY}"
      },
      "enabled": true
    },
    "fiber-ai-v3": {
      "type": "remote",
      "url": "https://mcp.fiber.ai/mcp/v3",
      "enabled": true
    },
    "fiber-ai-core": {
      "type": "remote",
      "url": "https://mcp.fiber.ai/mcp",
      "headers": {
        "x-api-key": "{env:FIBER_API_KEY}"
      },
      "enabled": true
    }
  }
}
```

If the file already has an `mcp` object, merge — do not overwrite existing servers.

If the file does not exist, create it with the full structure above.

## 3. Install the agent skills and personas

OpenCode reads skill-style instructions from `AGENTS.md` and/or `.opencode/AGENTS.md`, and persona subagents from `.opencode/agents/` (and `~/.config/opencode/agents/`). Do the following:

1. Fetch the list of skills from <https://github.com/fiber-ai/fiber-ai-plugin/tree/main/skills> and the personas from <https://github.com/fiber-ai/fiber-ai-plugin/tree/main/.opencode/agents>.
2. Clone the plugin repo into a sibling directory, or fetch the skill and persona Markdown files directly:

   ```bash
   git clone --depth 1 https://github.com/fiber-ai/fiber-ai-plugin .fiber-ai-plugin
   ```

3. Copy the persona subagents into the project's `.opencode/agents/` so they auto-load on matching intent:

   ```bash
   mkdir -p .opencode/agents
   cp .fiber-ai-plugin/.opencode/agents/*.md .opencode/agents/
   ```

   This installs seven personas: `ai-recruiter`, `ai-sdr`, `gtm-strategist`, `signal-scout`, `data-quality-auditor`, `product-engineer`, and `fiber-sde`. Users can also invoke them explicitly via `@ai-recruiter`, `@ai-sdr`, `@gtm-strategist`, `@signal-scout`, `@data-quality-auditor`, `@product-engineer`, or `@fiber-sde`.

4. Append a "Fiber AI skills and personas" section to the user's project `AGENTS.md` (create it if missing) that points to the cloned skills directory, the installed personas, and the canonical Fiber docs:

   ```markdown
   ## Fiber AI skills and personas

   Persona subagents (auto-load on matching intent, or invoke via `@<name>`):

   - `@ai-recruiter` - sourcing, JD to pipeline, GitHub + LinkedIn crosswalk, candidate reveal.
   - `@ai-sdr` - outbound list building, ICP definition, buyer identification, list pruning.
   - `@gtm-strategist` - pipeline math, market sizing, ABM strategy, scoping before commit.
   - `@signal-scout` - job-change / hiring / social / funding alerts on a seed list.
   - `@data-quality-auditor` - reproducible, pre-registered benchmarks vs competing data providers.
   - `@product-engineer` - real-time enrichment integrations, signup flows, identity resolution, API pipelines.
   - `@fiber-sde` - code generation, SDK usage, scripts, automations, and custom integrations with Fiber APIs.

   Workflow skills (called by personas, also usable directly):

   - "find companies like X" -> `.fiber-ai-plugin/skills/find-similar-companies/SKILL.md`
   - "enrich these LinkedIn URLs" -> `.fiber-ai-plugin/skills/enrich-linkedin-csv/SKILL.md`
   - "build a recruiting list" -> `.fiber-ai-plugin/skills/build-recruiting-audience/SKILL.md`
   - "reverse lookup these emails" -> `.fiber-ai-plugin/skills/expand-from-email-list/SKILL.md`
   - "enrich these GitHub handles" -> `.fiber-ai-plugin/skills/enrich-github-handles/SKILL.md`
   - "find VPs/CMOs/CTOs at ..." -> `.fiber-ai-plugin/skills/find-and-enrich-by-role/SKILL.md`
   - "track signals on these accounts" -> `.fiber-ai-plugin/skills/track-signals/SKILL.md`
   - "benchmark Fiber vs PDL/Apollo on this sample" -> `.fiber-ai-plugin/skills/benchmark-vs-competitor/SKILL.md`

   Canonical API docs: <https://api.fiber.ai/llms.txt>, <https://api.fiber.ai/ai-docs/index.md>.
   Per-operation docs: <https://api.fiber.ai/ai-docs/{operationId}.md>.
   ```

## 4. Verify

Restart OpenCode (or ask the user to) so the new MCP servers load. Then confirm with the user:

- "Ask me: 'find 5 fintech companies in NYC under 200 employees'"
- The agent should route to the `find-similar-companies` or `find-and-enrich-by-role` skill and call `companySearch` via the `fiber-ai-v2` MCP.

If the MCP connection fails:

- Confirm `FIBER_API_KEY` is exported (`echo $FIBER_API_KEY`).
- Confirm the URL reaches the server: `curl -sSI https://mcp.fiber.ai/mcp/v2`.
- Direct the user to <https://docs.fiber.ai/article/using-mcp-in-llms> for debugging.

## 5. Summary to print to the user

"Fiber AI is installed. Try: 'search for Series B fintech in NYC', 'enrich this LinkedIn URL', or invoke a persona directly with `@ai-recruiter build me a pipeline of staff iOS engineers`, `@ai-sdr build an outbound list of VP Marketing at Series B SaaS`, `@gtm-strategist help me scope an ABM program for infra tools`, `@signal-scout give me job-change alerts for these 50 accounts`, `@data-quality-auditor run a reproducible benchmark of Fiber vs PDL on my 200-row sample`, `@product-engineer integrate Fiber enrichment into my signup flow`, or `@fiber-sde write a script that enriches this CSV with emails`. I will route those through the Fiber MCP and the matching skill. Docs: <https://api.fiber.ai/llms.txt>."