About Tulip Agents
Tulip Agents is an MCP server published by tuliplabs-ai in the RAG category: the agent framework where the model never holds the trigger — every consequential action clears your policy first, waits for a human when it matters, and lands on a record you can verify. Build on it, or put it around the agent you already have.
Recent commit activity is the strongest signal that reported issues will get attention.
Install
From source
git clone https://github.com/tuliplabs-ai/tulip-agentsThis server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.
Ask AI
✦
Ask AI about Tulip Agents
Powered by Claude · Grounded in docs
I know everything about Tulip Agents. Ask me about installation, configuration, usage, or troubleshooting.
Security checks
6 checks have not been run yet.
Releases
v2.16.0v2.16.0 · 16 Sep 2026Added Computer use on each provider's native tool, gated like any other tool.** (#41) `computer_tool(session)` is one `Tool` named `computer` that drives a `BrowserSession` page by screenshot, pointer and keyboard. Anthropic sees Claude's…
v2.15.0v2.15.0 — voice, browser, Temporal, spend limits, graph history · 15 Sep 2026Realtime voice with gated tool calls, browser tools, durable runs on Temporal, spend as a policy input, and checkpoint history on graphs. [2.15.0] - 2026-09-15 Added Realtime voice sessions whose tool calls go through the gate.** (#18)…
v2.14.0v2.14.0 — approvals that hold, bind, and can be verified · 15 Sep 2026Approvals that hold across a restart, bind to the exact call, answer to approver rules, and leave a trail someone outside the runtime can verify — plus a token-counting context default, a spend budget, checkpoint history, a Docker sandbox…
v2.13.0v2.13.0 — code mode and deferred tools, both through the gate · 11 Sep 2026Everything since 2.12.3 — 2.13.0 is the first build on PyPI after the 2.12.4 and 2.12.5 entries, so all three sections ship here. [2.13.0] - 2026-09-10 Added Code mode: programmatic tool calling that still clears the gate.** (#176)…
v2.12.3v2.12.3 — what the hook replaced, what the tool failed with, who answered · 11 Sep 2026Three findings from a host that runs the SDK's whole hook surface — a streaming agent that slims bulky tool payloads before the model reads them. Fixed A hook's replaced tool result is actually used.** `AfterToolCallEvent` documents…