Hire Grant. Your COO and his ops team.

Grant is the COO for your software operations, an AI agent built by Since.dev. He leads a preset team of specialists who watch your GitHub Actions CI, AWS costs and runtime metrics, existing GitHub and AWS security findings, repository configuration and lifecycle, and dependency compatibility. Each reads its own sources and saves a brief. Ask Grant in Slack, the dashboard or your agent's MCP client: he consults the teammates your question needs, answers in one voice with the evidence and keeps track of what you decide. Free answers in the dashboard from saved results; paid plans consult the team live on your own model.

No. Grant and every specialist on his team are AI agents built by Since.dev, not people. They read the GitHub and AWS evidence you connect; supported repairs arrive as pull requests for your review, and nothing merges or deploys. Grant covers your software operations only: he doesn't hire, manage people or handle finance.

Free covers one project with a hosted AI allowance: one baseline and five follow-up check-ins per UTC month. Free includes Grant plus David, Maya, Liz and Priya in the dashboard, the last 30 days of history (older records are kept) and 250k hosted AI tokens a month. Grant answers from saved results and says as of when; he does not consult the team live. Pro adds Noah, Owen and Iris, Grant in Slack and MCP, full history and daily check-ins. Paid plans consult the team live in Slack, the dashboard and MCP, check in daily on Pro or every six hours on Studio, and also check in when GitHub reports a workflow run, a check, a push or a new alert. Paid work runs on your own Anthropic or OpenAI account, or AWS Bedrock from Studio, with any model ID the connection test marks ready for Grant.

Start on Home: choose GitHub or AWS and connect it. On GitHub, install the Since.dev app and pick a repository; it becomes your first project. On AWS, create a read-only role from the template we prepare, then name the project and choose what the team reads. On a paid plan, connect one AI provider. Then choose Approve and start check-in, and your team runs its first check-in right there. Only following public sources? Create a World State watch instead. Existing repositories and watches remain connected. Slack and signed webhooks are optional. Missing connections and unsupported resources appear as coverage gaps.

No. Check-ins and consults only read your connected GitHub and AWS evidence. Grant can record decisions and follow up, but cannot change AWS resources, rerun workflows or enable scanners. The existing repair flow can open draft pull requests for human review, and nothing merges on its own. Compatibility repairs keep their repository permissions, eligibility and repair settings. A cloud or security finding does not grant permission to write code or make production changes.

Once your Slack and Since identities are linked (automatically when your Slack email matches your Since account), ask with /grant or a mention in a channel: Grant answers publicly in that thread and discusses only the connection's projects, every project unless an owner narrowed them. DM Grant for a private, persistent chat over your own access. Grant reads only messages addressed to him (mentions, /grant and DMs with him), never broad channel history.

New paid repository connections default to Automatic draft PRs. Since automatically checks impact, prepares supported repairs, and opens draft PRs. Choose Monitor only to prepare repairs on demand. Repairs cover small, documented compatibility fixes in code and configuration. Routine version bumps and lockfile changes are outside that scope. If the evidence is insufficient, no PR is created. That does not prove your code is unaffected. Where GitHub doesn't offer draft PRs (private repositories on GitHub Free or Pro), the repair opens as a regular PR titled [Needs review]. CI runs in your repository. Since.dev never merges or deploys.

On Pro, connect Anthropic or OpenAI. Studio and Enterprise also support AWS Bedrock. Type any model ID your provider account can use, or pick a suggestion; the connection test checks it before saving, and Grant needs strict tool calling. One provider and model serves Grant, his team and repository AI. Your provider bills paid Grant and repository model usage directly. If the selected provider fails, work waits for recovery instead of silently switching accounts. Free repairs and World State watches include Since AI within their separate allowances. A watched subject counts once across its aspects and attached agents.