Skip to content

Database for agent-managed work. Private structured records agents can update across sessions.

The database for
agent-managed work.

Give AI agents one private place to read and update tasks, contacts, research, and feedback across sessions. Rowset is open source and self-hostable.

Start with work your agent already does.

You ask for the outcome. Your agent writes structured rows. Rowset keeps them ready for the next run.

Agent task board

Can replace

  • Linear
  • Asana
  • ClickUp
You ask
“Keep this project moving and tell me what is blocked.”
Agent writes
Tasks with owners, status, priority, and next steps.
Later
The next agent run resumes open work and updates the same task rows.

Feedback and bug tracker

Can replace

  • Zendesk
  • Intercom
  • Jira
You ask
“Collect feedback and bugs, then show me what needs attention.”
Agent writes
Reports with source, category, severity, status, and next step.
Later
New reports join the same queue, while resolved items stay searchable.

CRM

Can replace

  • HubSpot
  • Zoho
  • SAP
You ask
“Show me who needs a follow-up and why.”
Agent writes
Contacts, companies, relationship stage, last interaction, and next step.
Later
Future runs update the same contact history and follow-up queue.
See all workflow guides →

Open source. Self-hostable.

Run Rowset in our cloud or on your own infrastructure. Inspect the code, shape the roadmap, and keep deployment under your control.

One request becomes state the next run can use.

Follow one task from a chat request to an indexed row, a later agent handoff, and your review.

  1. You ask

    “Keep RFO-042 moving. Leave the next step where another agent can find it.”

  2. Agent A writes

    Agent A creates one row in agent_tasks, indexed by task_id.

    task_id
    RFO-042
    status
    Doing
    next_step
    Review mobile copy
    updated_by
    Agent A
  3. A later run reads

    Agent B retrieves the same indexed row for RFO-042 and continues from “Review mobile copy” instead of rebuilding the plan.

    Agent B updates that same row by task_id when the work is ready for review.

  4. You review

    In Rowset, you inspect the current row: task_id RFO-042, status Review, and updated_by Agent B.

    The record remains private and ready for the next run.

Build an agent task board →

A private database for agent-managed work.

Rowset stores structured rows with stable indexes, schema, instructions, and search. Agents access them through MCP, REST, or CLI—without direct credentials to your production database.

STORE
Typed, indexed rows
Stable business keys, generated IDs, descriptions, links, and flexible column types.
FIND
Exact and semantic search
Retrieve a known row or search by meaning with vector embeddings.
CONNECT
MCP, REST, and CLI
Use hosted tools from an agent or predictable endpoints from code.
MOVE
Five export formats
Take snapshots as CSV, JSONL, XLSX, SQLite, or Parquet.

Connect once. Then ask your agent to build.

  1. 01 / CONNECT

    Copy the setup prompt

    Give a trusted agent the Rowset setup prompt. It configures the best supported interface and verifies access with a bearer API key.

  2. 02 / RECOMMEND

    Start with useful structure

    Using already-authorized context, the agent recommends one project with one to three concrete datasets, then asks, “Would you like me to create that now?” If context is insufficient, it asks, “What are you working on right now?”

  3. 03 / DELEGATE

    Let it manage the rows

    The agent does not create the recommended project or datasets until you say yes. After you say yes, it creates or reuses the private resources, then reads, writes, searches, exports, and shares through authenticated tools. On a negative answer, create nothing.

Give your work one place to stay current.

Start with full product access for 7 days. Keep control with bearer-key authentication, explicit sharing, and exports whenever you want them.

Start your 7-day trial →