# Investboard
**Second brain for investing**
Four Apache-2.0 skills and reusable authored examples. The hosted app, MCP
backend and Premium data remain proprietary. [License scope](LICENSE.md).
Use four skills to understand portfolio exposures, review a proposed change against your own rules, work through a decision or clarify an investment thesis. Start with files you supply; connect your Investboard account when you need permitted account data.
## Try a rebalance review with real market data
**Your portfolio has drifted. What can available cash achieve before you decide
whether to sell?**
[Try the four-step rebalance demo](examples/portfolio-overlap/walkthrough.md#try-it).
Use observed market series for MSCI World and DAX with an explicitly illustrative
investor plan. Compare **buying underweights with new cash first**, then a smaller
contribution, selling equities and a cash-plus-sale hybrid. See the money required,
class weights and remaining drift, then draft a decision note.
[Try the connected rebalance workflow](examples/portfolio-overlap/connected-walkthrough.md):
read your portfolio and mandate, compare options using Investboard's native
calculation, and draft your reasoning. The hosted MCP now provides
`preview_rebalance` for this comparison. It requires authorized account access
and the applicable plan; your assistant must expose the new tool. The
supplied-file demo works independently.
1. **Install Portfolio Intelligence** using the [AI-agent installation options](#install-in-chatgpt-or-codex) or [individual skills](#install-individual-skills). The full walkthrough also uses Investment Strategy Check and Investment Decision-Making.
2. **Save and attach the four starting files** from the [walkthrough](examples/portfolio-overlap/walkthrough.md#try-it), then copy its first prompt into a new chat.
3. **Check the output** against the [worked comparison](examples/portfolio-overlap/README.md#compare-four-ways-to-proceed).
In this case, a €400 contribution restores the sample 90/10 asset-class targets gross
without selling. The unchanged portfolio is already inside its tolerance bands,
so the plan does not require a trade. Starting notionals, targets and cash
budgets are sample user inputs; the dated market observations are real.
Execution prices, fees, spreads and taxes still need their own inputs.
No Investboard account or broker connection is needed for the file-only demo.
Your AI provider’s account requirements, quota and fees apply. Worked answers
are reference calculations, not a model transcript or a claimed behavior pass.
If the toolkit helps, [give it a star on GitHub](https://github.com/Investboard/ai-toolkit).
## Four skills for better investment decisions
| Skill | What it helps you do | Example request |
| --- | --- | --- |
| [Investment Strategy Check](plugins/investboard/skills/investment-strategy-check/SKILL.md) | Connect your goals, allocation rules and constraints to a proposed change. | “Does moving ten percentage points from cash into equities fit my strategy?” |
| [Investment Decision-Making](plugins/investboard/skills/investment-decision-making/SKILL.md) | Work through choices, weigh evidence and alternatives, and revisit your reasoning. | “Help me work through this investment decision and the reasons to wait.” |
| [Investment Thesis](plugins/investboard/skills/investment-thesis/SKILL.md) | Clarify an investment idea, identify assumptions and decide what would change your mind. | “Help me turn these notes into an investment thesis I can review later.” |
| [Portfolio Intelligence](plugins/investboard/skills/portfolio-intelligence/SKILL.md) | Understand fund look-through, overlap, concentration, diversification and historical correlations. | “What do I actually own through these funds, and where do they overlap?” |
Skills work with evidence you supply. They do not sign you in or create access to live account data.
## Install in ChatGPT or Codex
[Open the published Investboard plugin](https://chatgpt.com/plugins/plugin_asdk_app_6acd243688888191a3eda79bea1ca0d0). In ChatGPT, install it and choose **Try in chat**. In Codex, search for Investboard and choose its public directory entry. Local marketplace and personal test entries are separate copies. Start a new conversation with your files, or authorize your Investboard account when you need connected tools. Reuse an existing connection; workspace policies may require administrator approval.
## Install in Claude Code
Add the Investboard marketplace, then install its plugin:
```text
/plugin marketplace add Investboard/ai-toolkit
/plugin install investboard@investboard
```
Start a new conversation. The plugin includes all four skills and the Investboard MCP connection. Complete OAuth when you want account access; review the requested permissions before connecting.
## Install individual skills
For native client setup, see [VS Code / GitHub Copilot](clients/vscode/README.md)
and [Gemini CLI](clients/gemini/README.md).
Use a skills-compatible assistant and choose the skills you want:
```sh
npx skills add Investboard/ai-toolkit
```
For Claude chat, download an [individual skill ZIP](https://github.com/Investboard/ai-toolkit/releases/latest), enable code execution and file creation, then open Customize → Skills → Add skill → Upload skill. Enable the skill and start a new conversation.
The repository also includes a portable Agent Plugins manifest and a Codex marketplace catalog. Follow the [installation guide](https://investboard.com/product/ai-toolkit#install) for your client. A client must support the relevant skill or plugin format; account and workspace requirements vary by provider.
## Analyze holdings without an account
The same MCP endpoint also provides `analyze_holdings` and
`stress_test_holdings` for holdings you supply in the conversation. These two
tools need no Investboard sign-in and read or save nothing in an account.
Supply your actual ISINs, weights or values and currency; missing reference data
is reported as partial or unresolved. Historical stress results replay the past,
not a forecast. Your assistant provider's requirements and usage limits apply.
## Connect your Investboard account
All four skills can work with files you supply without an Investboard subscription.
**The full connected experience requires Investboard Premium:** portfolio analytics
and reading saved check history, investment theses and journal entries. Free accounts
include one connection for policy and status, checks and recording decisions you state.
Your AI assistant provider's subscription is separate from Investboard Premium.
Connecting an assistant does not start a subscription or change your plan.
[Create an Investboard account](https://investboard.com/sign-up) and complete setup,
then return to your assistant to connect. [Compare Free and Premium](https://investboard.com/pricing).
[Choose your assistant and install Investboard](https://investboard.com/product/ai-toolkit#install)
Use the direct remote MCP endpoint with OAuth:
```text
https://app.investboard.com/mcp
```
Sign in to your own Investboard account. If you already have this connection, use it instead of adding a duplicate. Portfolio access requires `portfolio:read` consent and enabled portfolio sharing in Investboard.
Start with: “Read my Investboard policy and explain its date and rules. Do not run a check or record a decision yet.”
### Connected capabilities
| Tools | Capability |
| --- | --- |
| `get_policy`, `get_status` | Read your investment mandate and its current standing. |
| `check_action`, `record_decision` | Save a hypothetical check and record an outcome you explicitly state. |
| `get_reflection_history`, `get_own_record`, `get_journal` | Retrieve permitted checks, theses and journal entries. |
| `get_portfolio_holdings`, `get_portfolio_analysis`, `get_portfolio_correlations` | Read holdings, fund exposures and historical portfolio analytics. |
| `get_performance_explanation`, `preview_rebalance`, `get_decision_outcomes` | Explain whole-portfolio performance, compare contribution and sale alternatives, and revisit recorded decision outcomes. The comparison saves no check or plan. |
| `get_review_changes`, `prepare_review`, `get_prepared_review`, `get_approved_review` | Read changes since a review, prepare the investor's stated reasoning for approval in Investboard, and retrieve prepared or approved reviews. Preparing a review does not approve it. |
| `list_stress_scenarios`, `start_stress_test`, `get_stress_test` | Replay a historical crisis over today's holdings. Starting a run saves it in stress-test history; it changes no holdings. |
| `prepare_portfolio_import`, `get_portfolio_import` | Stage supplied holdings and read the receipt. The investor must confirm the import in Investboard before anything enters their portfolio. |
The hosted tool catalog updates independently of the four packaged skills.
Published directories may expose new tools only after their checks finish. If a
tool is missing, follow your assistant's refresh flow and start a new
conversation; keep your existing connection.
The full connected workflow requires **Investboard Premium**. Active Free accounts support one connection and policy, status, check and decision-recording tools. Premium adds unlimited connections, supported check history, thesis and journal reads, and portfolio analytics. Your assistant provider has separate requirements.
Checks and stated decisions are saved in your account. Imports remain pending
until you confirm them; prepared portfolio reviews require your approval in
Investboard. Account stress tests create history records. The tools do not
submit orders, transfer money, change your mandate, or edit theses and journal
reviews. Portfolio reads can include holding values when authorized; they do not
return brokerage credentials or account numbers. Keep evidence dates, currencies
and coverage with any interpretation. Historical correlations do not guarantee
future diversification.
Manage permissions, portfolio sharing and revocation in Investboard’s assistant settings. Revocation stops future access; it does not recall information already shared.
## Downloads and updates
[Download the latest release](https://github.com/Investboard/ai-toolkit/releases/latest) for the complete plugin, marketplace bundle or individual skills. Release assets include SHA-256 checksums. The website also hosts versioned downloads. A newer repository commit does not mean its ZIP release is already published. Download versions are separate from the published OpenAI plugin version.
Update the installed skills or plugin through your client’s update flow. Update individual ZIP installations by uploading the newer version. Package versions are separate from the MCP server’s Registry version.
## Evaluate the skills
The [offline evaluation harness](evals/README.md) includes 24 synthetic scenarios for the four skills. Validate the cases locally or grade saved agent traces with explicit human review. The harness makes no model calls; synthetic fixtures and missing runs do not count as behavior passes.
## License and help
The four skills and Investboard-authored examples are **open source under
Apache-2.0**. You can use, adapt and redistribute them under that license.
Provider observations in the example are excluded from the grant.
The app, hosted MCP backend, Premium data, plugin wrapper and branding remain
proprietary. Hosted-service use is subject to the [Investboard terms](https://investboard.com/terms).
See the [license scope](LICENSE.md) and [example data exclusions](examples/THIRD_PARTY_DATA.md).
[Installation guide](https://investboard.com/product/ai-toolkit#install) · [Help](https://investboard.com/help) · [Privacy](https://investboard.com/privacy) · [Terms](https://investboard.com/terms)