> ## Documentation Index
> Fetch the complete documentation index at: https://docs.timepointai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Timepoint Conductor

> The natural-language operator of the whole Timepoint stack — find money, run a simulation, render a moment, explore the graph, all in plain language. Available over REST and MCP.

# Timepoint Conductor

**Operate the entire Timepoint stack in plain language.** Ask the Conductor to find money, run a simulation, render a moment, or explore the causal graph — it picks the right services, runs and monitors the work, recovers when something fails, and reports the credits it spends as it goes.

The Conductor is the **authenticated home of [app.timepointai.com](https://app.timepointai.com)** and is available over both the **REST API** and **MCP**, so a person in the web app and an agent driving it programmatically share the same operator.

<CardGroup cols={2}>
  <Card title="Agent / MCP Usage" icon="plug" href="/api-reference/conductor">
    Drive the whole stack from one endpoint — the MCP `converse` tool, auth, and an example request/response.
  </Card>

  <Card title="The Find Money Flow" icon="magnifying-glass-dollar" href="/products/flash">
    The Conductor can run a full Find Money flow end to end, handling the intermediate screening itself.
  </Card>
</CardGroup>

<Note>
  The Conductor calls every service through the [API Gateway](/api-reference/overview) (`api.timepointai.com`), so authentication, credits, and spend metering stay handled in one place. You authenticate the Conductor with your normal Timepoint token — the same credentials you use everywhere else.
</Note>

## What you can ask it

You talk to the Conductor the way you'd brief a capable operator. For example:

* *"Find me \$50k in grants for a regional arts nonprofit by September."* — it sources opportunities, runs the quick simulations, screens them down, and runs the deep simulation, then hands back the strongest leads.
* *"Run a Pro simulation on the top three of those and compare the strategies."*
* *"Render the moment AlphaGo played Move 37 against Lee Sedol."*
* *"Show me what's connected to the fall of Constantinople in the graph."*
* *"How many credits do I have left, and what did that last run cost?"*

It routes each request to the right service (Flash, Pro, Clockchain, Find Money, Billing, TDF, and more), runs long jobs and polls them to completion, and folds the results back together — imagery from Flash, simulation data from Pro, graph structure from Clockchain — into one reply.

## How it behaves

**It reports spend as it works.** The Conductor is credit-aware. Every time it calls a tool it emits a short *spend chirp* — what it did, what it cost, and the running total — so you always know what a request is consuming. It tracks its own usage separately and never charges you outside the normal credit ledger.

**It recovers from failure.** If a step errors, the Conductor doesn't give up — it re-approaches: retrying with adjusted parameters, falling back to a cheaper path, or asking you how to proceed. Long-running Pro jobs are polled with backoff; Find Money runs are driven through their stages, with the Conductor making the intermediate selection itself when you've asked it to.

**It works on your own data — after a check.** You can ask it to build on entities you've created before. The Conductor confirms you own (or have been granted access to) an entity before it uses it.

## Tuning your Conductor

You shape how your Conductor behaves through a small set of fixed options (no free-form configuration — the choices are a locked menu). Your selections are saved per account and applied to every conversation.

| Option             | Choices                                                                     | What it controls                                                                                            |
| ------------------ | --------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- |
| **Autonomy**       | `make choices for me` · `stop at inflection points` · `confirm every spend` | How much the Conductor decides on its own versus checking in with you — especially before spending credits. |
| **Verbosity**      | `terse` · `normal` · `narrate chirps`                                       | How much the Conductor explains as it works, including how loudly it narrates spend chirps.                 |
| **Spend ceiling**  | `ask always` · `50` · `200` · `1000` credits                                | A per-request credit budget. Above the ceiling, the Conductor pauses to confirm before continuing.          |
| **Default render** | `text` · `rich`                                                             | Whether replies come back as plain text or fold in imagery, data, and TDF-native graphics.                  |

For example, set **autonomy** to `confirm every spend` and **spend ceiling** to `ask always` for maximum control, or `make choices for me` with a `200`-credit ceiling to let it run multi-step work uninterrupted up to a budget you're comfortable with.

## Available over REST and MCP

The Conductor speaks two ways:

* **REST** — `POST /api/v1/conductor/messages` with your Timepoint token. This is what powers the chat at `app.timepointai.com`.
* **MCP** — a Model Context Protocol endpoint at `conductor.timepointai.com/mcp` exposing a `converse` tool, so any MCP-capable agent (including Claude) can drive the full Timepoint stack from a single endpoint.

See [Agent / MCP Usage](/api-reference/conductor) for the developer-facing details.

## Where it lives

| Surface                  | URL                                                                |
| ------------------------ | ------------------------------------------------------------------ |
| Web chat (authenticated) | `app.timepointai.com`                                              |
| REST API                 | `POST https://conductor.timepointai.com/api/v1/conductor/messages` |
| MCP endpoint             | `https://conductor.timepointai.com/mcp`                            |