Project memory for cloud coding agents
- MCP
- API
- Web
Today agent knowledge is a silo on every team. Briefings stay local, the next agent rebuilds them, and you pay for those tokens again and again. Supervela is the shared project memory that your team retrieves against, grounded in what you already wrote, what was decided in meetings, and the code in the repo.
Cap requests per API key, return 429 with a retry hint, and keep the queue worker idempotent. Acceptance criteria came from the linked ticket.
- KAN-142 · Jira
- Ingest limits · Confluence
- Platform decisions · Meeting
How Supervela works
One store per workspace and project. What the team already wrote, what was decided in meetings, and the code in the repo — one place. Browse, search, and cite the same context in work orders.
Is this story, task, or bug AI-ready? We check project knowledge and the repo when it is connected, so you only send work a cloud agent can finish.
Send a work order to a cloud agent. Pull requests come back in one list.
Chat with the knowledge store in the console, or point any MCP-capable editor at our hosted MCP server — both read the same project memory. Available on Teams, Enterprise, and Dedicated.
Individual and above. Workspace admins create API keys in Settings. Query project memory from n8n, Make, or your own scripts.
Teams and above. Push decisions and actions into the store from the tools you already run.
Keyword and semantic search over the same store, with ranked results. Knowledge items link in a graph that can be traversed.
Workspaces run in AWS on a serverless API, Aurora Postgres + pgvector, and queue workers. Nothing to install and no database on a developer machine.
Each workspace is isolated in Postgres with row-level security, Cognito sign-in, and membership scoped to the projects a person works on.
Plans
One knowledge store. Connect Jira, Confluence, GitHub, and Drive meeting folders. Run cloud agents.
30-day trial.
- Knowledge base
- Claude / Cursor cloud agent integration
- Atlassian / GitHub / Google Drive integration
- Single seat, single project
- 5 AI jobs a day, 40 a month
- 1 transcript upload a day
£15 / month
For one person who uses it every day.
- Semantic search
- Read API
- 1 seat
- 2 projects
- 50 AI jobs a day, 200 a month
- 5 transcript uploads a day
£25 / seat / month
One store for a small team and the agents they send work to.
- Chat with the knowledge store
- MCP
- Ingest API
- 2–5 seats
- 10 projects
- 100 AI jobs a day, 400 a month
- 15 transcript uploads a day
Need more than 5 seats or 10 projects? Contact us for Enterprise and Dedicated plans.