reflectionbeam.org
en

Agents

Coding agent configs

Choose your agent and options to get the install command, config files and a check that the agent really uses Beam. Based on Reflection's coding agent quickstart.

OpenCode
Agent
Default reasoning effort
Config location
Open-source terminal agent, tested by Reflection with OpenCode 1.18.34. interleaved.field tells it to send reasoning_content back. Keep "model" in the config, or OpenCode may start on another provider's model.

  1. Install

    bash
    curl -fsSL https://opencode.ai/install | bash
# or: npm install -g opencode-ai

  2. Configure

    Terminal
    export REFLECTION_API_KEY="<your API key>"

# Check the key: the list should include Beam-501B-A23B
curl https://api.reflection.ai/openai/v1/models -H "Authorization: Bearer $REFLECTION_API_KEY"
    opencode.json
    {
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "reflection": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Reflection",
      "options": {
        "baseURL": "https://api.reflection.ai/openai/v1",
        "apiKey": "{env:REFLECTION_API_KEY}"
      },
      "models": {
        "Beam-501B-A23B": {
          "name": "Beam-501B-A23B",
          "modalities": {
            "input": [
              "text"
            ],
            "output": [
              "text"
            ]
          },
          "limit": {
            "context": 262144,
            "output": 131072
          },
          "reasoning": true,
          "interleaved": {
            "field": "reasoning_content"
          },
          "variants": {
            "xhigh": {
              "reasoningEffort": "xhigh"
            },
            "max": {
              "reasoningEffort": "max"
            }
          }
        }
      }
    }
  },
  "model": "reflection/Beam-501B-A23B",
  "permission": {
    "bash": "ask",
    "edit": "ask"
  }
}

  3. Check the model

    bash
    opencode models reflection
# prints: reflection/Beam-501B-A23B

  4. Run

    bash
    cd your-project
opencode
# one-shot: opencode run --variant high "<prompt>" </dev/null

Known issues

400: Missing required parameter reasoning_content

The session sent back an assistant step without its reasoning. Start a new session (/new, or opencode run without -c). In Hermes, check that model.reasoning_echo is true.

The agent uses a different model

Check the model name the agent prints at start. In Mirror run /model reflection; in Pi and OpenCode check the default provider and model in the config.

404 or model not found

The base URL must end in /openai/v1 with nothing after it, and the model ID must be exactly Beam-501B-A23B.

Long pauses before output

Beam reasons before it answers, and large tasks can take a while. On repeated 502 or 503 errors, retry the task.

Source: developers.reflection.ai/coding-agents-quickstart