HiveBase
HiveBase
API + MCP

One shared context, callable from every AI app you already use.

One hosted MCP endpoint and 68 agent-ready tools. Your agents read your company's context before they work and write results back — with provenance and one-click undo. Install once, works everywhere.

Read the docsMCP connector install →
0
agent-ready tools
across 9 domains
Free
reads — always
0 credits for all context.search, get_entity
OAuth 2.1
hosted auth
with dynamic client registration
1-click
undo on any write
provenance-tagged, reversible
INSTALL OVER MCP
Claude Code
claude mcp add hivebase \
  --transport http \
  https://mcp.hivebase.ai/v1
Cursor
# .cursor/mcp.json
{
  "mcpServers": {
    "hivebase": {
      "url": "https://mcp.hivebase.ai/v1",
      "transport": "http"
    }
  }
}
Codex
# codex.yaml
mcp_servers:
  - name: hivebase
    url: https://mcp.hivebase.ai/v1
    transport: http

Authentication via OAuth 2.1 on first connection. Installs into Claude Code CLI, Conductor, Codex, Cursor, and any host that supports Streamable HTTP transport.

READ · DISPATCH · WRITE + UNDO

From terminal to cloud task — and back.

Read context for free, dispatch a cloud task from any AI app, write results back with provenance — and undo with a single call. All from the same MCP session your agent is already using.

tasks.dispatch is in preview — shown as aspirational in the demo

HIVEBASE MCP
READDISPATCHWRITE + UNDO
THE SAME ANSWER, EVERYWHERE

The same context, inside Claude Code, Cursor, and Codex.

Context flows in zero-effort from everywhere. The exact same answer comes out everywhere — including inside the AI apps you already work in.

  • Parity-enforced — identical input, identical answer, everywhere
  • One hosted endpoint, 68 tools, install once over MCP
  • Agents propose, humans approve — every write provenance-tagged
CONTEXT READS FROM HIVEBASE
QUERY
PARITY
Claude Code
Cursor
$ claude mcp add hivebase --transport http \
  https://mcp.hivebase.ai/v1
68 TOOLS ACROSS 9 DOMAINS
Brain / Context
query, search, get_entity, create_entity, update_entity
Reads are always 0 credits
Tasks
create_task, list_tasks, update_task, dispatch_code_agent
Dispatch from any AI app
Strategy
get_overview, list_competitors, get_intel_feed
Accounts
get_health, list_at_risk, create_note
Calendar + Meetings
list_upcoming, get_prep, add_note
Decisions
record_decision, list_decisions, get_decision
Provenance-tagged, reversible
FYI
get_brief, dismiss_item, mark_done
Notifications
create_action_item, complete_action_item
Integrations
list_connected, trigger_sync
Agents propose, humans approve
Any tool that writes (create_entity, record_decision, update_task) carries requires_confirmation: true for oauth tokens. You approve or dismiss from the AI app or HiveBase UI.
Reads are always free
context.search, get_entity, query, and all read tools cost 0 credits. Your agents can read as much as they need — you pay only when work happens.
Provenance on every write
Every write is tagged with the agent identity, timestamp, and reasoning. One-click undo is available for 7 days on any write made via MCP.
One shared context, callable from everywhere.
Install the MCP connector and your agents inherit your company's context before they work. Reads free. Writes provenance-tagged and reversible.
Read the docs