D
DevcoreAI
Download
BYOK model setup

Connect your AI models to DevCoreAI safely.

DevCoreAI currently works as BYOK: you connect your own provider keys, your provider bills model usage directly, and DevCoreAI gives you the agent workspace, approvals, settings, platform limits, and dashboard experience.

API configuration reference Understand plans

Provider billing is separate

Your provider charges model tokens directly. DevCoreAI plans cover platform access, automation, team features, and governance.

Keys stay private

Store provider keys only in DevCoreAI settings. Do not paste secrets into prompts, docs, screenshots, or issue comments.

Use the right model

Match model strength to risk. Use stronger models for architecture and security, faster models for small edits.

Fast setup checklist

  1. 1Open DevCoreAI Settings.
  2. 2Go to API Configuration.
  3. 3Choose the provider you want to use.
  4. 4Paste your provider API key or local base URL.
  5. 5Select a model your account can access.
  6. 6Choose reasoning effort appropriate for the task.
  7. 7Save settings.
  8. 8Run a tiny test prompt before approving file edits or terminal commands.

Provider-specific setup

OpenAI

Fast coding, general reasoning, and broad tool support.

  1. 1Create or open your OpenAI platform account.
  2. 2Create an API key with billing enabled.
  3. 3Open DevCoreAI Settings > API Configuration.
  4. 4Choose OpenAI as the provider.
  5. 5Paste the API key into the provider key field.
  6. 6Select the model your account can access.
  7. 7Save settings and run a small test prompt before a real task.

Anthropic

Large refactors, architecture planning, careful code review, and long context workflows.

  1. 1Create or open your Anthropic Console account.
  2. 2Create an API key and confirm billing is active.
  3. 3Open DevCoreAI Settings > API Configuration.
  4. 4Choose Anthropic as the provider.
  5. 5Paste the API key and select a Claude model available to your account.
  6. 6Use a stronger model for Plan mode when the task touches many files.
  7. 7Run a short test request and confirm the response streams in the chat panel.

Google Gemini

Long-context analysis, broad repository reading, and cost-sensitive workflows.

  1. 1Create a Google AI Studio or Google Cloud API key.
  2. 2Confirm the selected Gemini model is available in your account and region.
  3. 3Open DevCoreAI Settings > API Configuration.
  4. 4Choose Google or Gemini as the provider.
  5. 5Paste the key, select the model, and save.
  6. 6Use a small file-reading task first to confirm context and response behavior.

OpenRouter

Trying many providers from one API key and quickly switching between models.

  1. 1Create an OpenRouter account.
  2. 2Add credits or enable billing.
  3. 3Create an API key.
  4. 4Open DevCoreAI Settings > API Configuration.
  5. 5Choose OpenRouter as the provider.
  6. 6Paste the key and select the exact provider/model route you want.
  7. 7Check OpenRouter logs if a model is blocked, rate-limited, or unavailable.

Local models

Private experimentation, offline-friendly workflows, and cost control for simpler tasks.

  1. 1Install and run a local model server such as Ollama or LM Studio.
  2. 2Download a model that fits your hardware.
  3. 3Confirm the local server URL is reachable from your machine.
  4. 4Open DevCoreAI Settings > API Configuration.
  5. 5Choose the local provider option.
  6. 6Enter the local base URL and model name.
  7. 7Run a small prompt before using file edits or terminal actions.

What every model setting means

SettingHow it works
ProviderThe service DevCoreAI sends model requests to. Change this when switching from one model company or router to another.
API keyYour provider credential. Keep it private, rotate it if exposed, and never paste it into normal chat messages.
Base URLUsed for local models, compatible gateways, self-hosted proxies, or enterprise model routers.
ModelControls output quality, context size, speed, and provider cost. Use stronger models for high-risk work.
Reasoning effortControls model depth when supported. Higher effort is useful for architecture, security, debugging, and multi-file refactors.
Plan/Act model splitLets you use one model for planning and another for implementation. This can reduce cost while keeping analysis quality high.

Choose the right model for the task

You are fixing one small bug

Use a fast, cost-efficient model in Act mode.

You are changing auth, billing, database, or security code

Use a stronger reasoning model, start in Plan mode, and keep approvals manual.

You need to inspect many files

Use a long-context model and ask DevCoreAI to analyze before editing.

You want to compare many model providers

Use OpenRouter or a compatible router, then pin the best model per workspace.

You need private local experimentation

Use Ollama or LM Studio, but verify model quality before production edits.

Enterprise policy and user settings

In team or enterprise environments, admins may use DevCoreAI remote policy to limit providers, hide unsupported models, require manual approvals, or disable risky integrations. User and workspace settings still appear inside the IDE, but some values may be locked by organization policy.

If a provider or model does not appear in settings, check whether your organization policy hides it before assuming the install is broken.

Troubleshooting model connections

Invalid API key

Create a new key, paste it into API Configuration, save, and confirm the provider account has billing or credits enabled.

Model not found

Check the exact provider model ID, account access, region availability, and whether an admin policy hides that model.

Rate limit or quota error

Wait for quota reset, upgrade provider limits, choose a lighter model, or reduce parallel agent activity.

Local model does not respond

Confirm the local server is running, base URL is correct, model is downloaded, and firewall rules allow local access.

Provider cost is higher than expected

Use Plan/Act split, lower reasoning effort for simple work, avoid huge context uploads, and review provider billing logs.

Enterprise policy blocks a provider

Ask your workspace or organization admin to review DevCoreAI remote policy. User settings may be read-only when policy is enforced.

Model connected?

Start in Plan mode for the first real task. Ask DevCoreAI to inspect the codebase and propose a plan before approving edits.

Continue to Plan and Act