Investboard …
KnowledgeStrategy & portfolio
How to use AI in Investboard: A practical guide
Bring your rules into the conversation with investing.md, MCP, plugin skills and Premium portfolio intelligence. Follow the steps and try the example prompts. Bring your rules, check an idea and preserve your reasoning.
In short
Use AI in Investboard with your own mandate as the reference. Built-in AI explains available portfolio data. investing.md carries your rules as a dated file; MCP gives supported AI agents permission to retrieve records and check a proposal. Record your thesis and falsifier, then review observations before revisiting the original reasoning. You retain every investment decision.
Built-in AI and an external MCP connection have different tools and data access.
investing.md is a dated snapshot; saved instructions accompany MCP policy reads.
Additional key points
A check saves the proposal and its evidence, but executes no order.
A later review starts with your observation and preserves your earlier reasoning.
In short
Use AI in Investboard with your own mandate as the reference. Built-in AI explains available portfolio data. investing.md carries your rules as a dated file; MCP gives supported AI agents permission to retrieve records and check a proposal. Record your thesis and falsifier, then review observations before revisiting the original reasoning. You retain every investment decision.
Built-in AI and an external MCP connection have different tools and data access.
investing.md is a dated snapshot; saved instructions accompany MCP policy reads.
A check saves the proposal and its evidence, but executes no order.
A later review starts with your observation and preserves your earlier reasoning.
A useful AI investing conversation starts with the rules you chose and the reasons you recorded. This guide shows how to bring both into Investboard's built-in AI or an external AI agent, check a proposal, and return to your reasoning when it is time to review.
AI investing in two pages. Premium market data for your research. Your strategy and investing context for agentic systems, through investing.md, skills and MCP.
Download the English overview (PDF, 2 pages)Choose your starting point
Use the route that fits the task in front of you. You can begin with a file and connect an AI agent later.
For a one-time assessment without an account or ongoing portfolio storage, start with the personal portfolio report from broker CSVs. For ongoing tracking, use the web and mobile platform. The complete connected AI workflow requires Premium. You can also use the skills with files you supply; active Free accounts support the limited MCP access described below.
| Your task | Where to start |
|---|---|
| Understand available portfolio data | Built-in AI in Investboard |
| Give another AI agent your written rules | Export your personal investing.md file |
| Read current policy and save a proposal check | Connect a compatible AI agent through MCP |
| Preserve your reasoning and review it later | Research, theses and the Reflection journal |
Investboard AI has its own portfolio tools. The external Investboard MCP connection provides tools for your mandate, saved reasoning and portfolio analysis. This guide focuses on those workflows. Portfolio reads require portfolio:read, Premium and enabled portfolio sharing. An uploaded file gives an AI agent only the text you share.
Investboard does not place, change or transmit orders. If your AI agent also has a separate broker connection, that connection has its own permissions. Keep each proposed action attached to the tool that would perform it.
The Premium offer: your rules, your data, your reasoning
With Investboard Premium, a permitted AI agent can put a proposal beside your imported holdings, supported fund exposures, historical correlations and saved reasoning. Premium includes unlimited MCP connections, supported check-history, thesis and journal retrieval, and portfolio analytics.
The plugin package brings four named skills to that workflow: Investment Strategy Check, Investment Decision-Making, Investment Thesis and Portfolio Intelligence. They guide the process from clarifying a question to examining evidence and reviewing your own reasoning. Setup depends on the AI agent you use. Skills do not grant account access or change your subscription.
For portfolio questions, the value is evidence you can inspect: underlying fund exposures, overlap, concentration and historical correlations, alongside returned sources, dates and coverage. Portfolio tools need portfolio:read, Premium and enabled portfolio sharing. Your AI agent provider may have separate requirements.
Give your AI the portfolio behind your rules
Connect portfolio intelligence, your investment policy and your decision history with Investboard Premium.
See Premium plansInvestboard Skills
For the full connected workflow, you need a separate Investboard Premium subscription. Create your Investboard account, complete setup and return to your AI agent to connect. Compare Free and Premium. Your AI provider's subscription is separate; all four skills can also work with files you supply.
Choose your AI agent in the AI Toolkit installer. ChatGPT and Codex can install the published Investboard plugin. Claude Code can use the public GitHub marketplace. Claude chat uses the individual skill ZIPs. Local marketplace ZIPs remain an alternative; a plugin ZIP is different from an individual skill ZIP.
Using ChatGPT? Read Investboard in ChatGPT: a second brain for investing for the plugin's four skills, connected Second Brain features and the Free/Premium distinction.
After installation, start a new conversation and choose the skill that matches your question. Supply your investing.md, notes or report, or authorize an MCP connection for supported account data.
The toolkit includes a first message for each skill and help with uploads, connections and missing data. Keep local marketplace folders after installation. ZIP installations are snapshots; return to the toolkit for updates. Removing a skill is separate from revoking account access under AI agents → Connections.
- Investment Strategy Check: Check a proposed buy, sell or rebalance against your own goals, allocation ranges and limits. Review the rules touched, data coverage and missing evidence.
- Investment Decision-Making: Compare the alternatives you bring, make trade-offs explicit, and record your stated reasoning when you ask. Revisit a decision without confusing the outcome with the quality of the reasoning.
- Investment Thesis: Structure or review your investment case around assumptions, time horizon, risks and the evidence that would change your view. Save a thesis through Investboard.
- Portfolio Intelligence: Examine underlying fund holdings, overlap, concentration and historical correlations using authorized analytics or a report you provide. Keep dates and coverage attached to the explanation.
A supplied file is a snapshot. Reading account data or recording a decision requires the appropriate MCP permissions and plan. The skills guide the analysis; you make the decision.
Before you start: check your mandate and data
Record your purpose, horizon, risk tolerance, allocation and review rules in your mandate. Complete the account setup and portfolio review presented by the product. Then verify the holdings, quantities, currencies and data dates in your portfolio. The portfolio-check guide explains that preparation.
If two screens disagree on a holding's value or the data is incomplete, resolve the discrepancy before relying on a numerical check. A fresh tool response can still contain older underlying evidence.
Active Free accounts support one MCP connection and the core policy, status, check and decision-recording tools. Premium adds unlimited connections, supported history, thesis and journal retrieval, and portfolio analytics. Your AI agent also needs the matching permissions; a subscription alone grants no connection access. See Free and Premium.
Open the free investing.md workbook, or read the format and examples. Your file supplies policy context; account access requires a separate connection.
Step 1: Ask built-in AI about your data
Open Investboard AI in Investboard's app toolbar. If prompted, review the AI processing information and choose whether to allow it. Start with a question about data already available in your account.
Explain my current allocation. Separate measured facts from interpretation, give the data dates, and name any missing coverage. Tell me which displayed figures support your explanation.
Read the answer beside the relevant portfolio view. Check which scope and date it describes. If the figures do not match, ask for clarification rather than treating a fluent explanation as evidence.
Use AI at Investboard for Investboard AI's scope and processing details. External MCP setup is a separate step below.
Step 2: Take your mandate with you as investing.md
For a conversation without an account connection:
- Open Strategy → Overview → Brief.
- Find Export your strategy for your AI and choose Download or Copy.
- Check the file's
composed_atdate. Share the file or paste its contents into the AI agent you choose.
The signed mandate also offers Download investing.md. The welcome certificate's Save as a text file downloads the same policy format; Copy link copies an owner-only page link, not the policy contents.
To add your own instructions, open AI agents → Instructions. Write Your additions and choose Save additions. The structured mandate takes precedence over this separately attributed text. Saved file lets you copy or download the document. An unsaved preview is labeled as a draft; a connected AI agent receives saved additions, not unfinished editor changes.
Try this with your exported file:
Read this investing.md as my investment policy. State its composition date, then summarize my purpose, allocation bands, review rules and cooling-off terms. Quote my own additions separately. Ask me to clarify anything missing. Do not infer current holdings or permission to access my account from this file.
Completion check: the AI agent identifies the date and distinguishes generated mandate rules from your additions. If you later amend the mandate, export and share a replacement. The file stays a snapshot; it does not enforce model behavior or keep itself current. Read about the investing.md format.
Step 3: Connect an AI agent through MCP
MCP lets a compatible AI agent call Investboard's supported tools after you authorize access.
- In Investboard, open AI agents → Connections and choose your AI agent.
- Install the published Investboard plugin where available, or follow your AI agent's setup steps in the toolkit installer. For a manual remote MCP connection, use the server address that Investboard shows. Account connections use OAuth sign-in.
- Sign in to the Investboard account you want to connect. Review the requested permissions before approving them.
- Enable or invoke the connection in your conversation and ask it to read your policy.
For Claude, begin in Customize → Connectors and add a custom connector. Use OAuth sign-in, then enable the connector in the conversation. Team setup may require an owner. Follow Claude's current instructions.
For ChatGPT, open the published Investboard listing, install the plugin and choose Try in chat. The package includes all four skills and the account connection. Connect your Investboard account when you need account data, then invoke Investboard with @. Workspace policies may require administrator approval. A custom MCP connection is an optional alternative, following OpenAI's instructions; reuse an existing connection rather than adding a duplicate.
For Codex, search for Investboard in its plugin directory and choose its public directory entry. Local marketplace and personal test entries are separate copies. Start a new conversation and use the skills with your files, or authorize account tools when needed. ChatGPT and Codex share the public Investboard listing.
For Claude Code, add the public Investboard marketplace and install the plugin:
claude plugin marketplace add Investboard/ai-toolkit
claude plugin install investboard@investboard
Start a new session. Complete OAuth when you want the connected account tools. The installer also documents a local ZIP alternative.
Other clients need remote MCP with Streamable HTTP and OAuth support. Choose your client in the toolkit installer for its setup steps. Developers running TradingAgents locally can use the separately published tradingagents-investboard connector, also available on PyPI. Follow that package's own requirements and authorization instructions; it is a separate integration from the four skills and the remote MCP connection.
Your first message can be:
Use Investboard's get_policy to read my current mandate and saved instructions. Summarize the rules with the returned composition date. Do not run a check yet. Ask me which change I want to examine.
Completion check: inspect the tool response, not just the AI agent's claim that it is connected. It should contain your policy. The connection appears in Investboard after the AI agent first calls it. A policy read alone creates no check in Activity.
Step 4: Check a proposed change and record your decision
Describe the change completely, including its funding. A purchase funded from settled cash needs both the purchase and the cash reduction. A purchase funded by selling another holding needs both legs. Money added from outside the portfolio should be stated as new money.
For a hypothetical proposal, try:
Check this hypothetical change against my mandate: use €1,000 of settled cash to buy €1,000 of equities within the existing portfolio. Include the cash reduction and equity purchase as separate legs. Show the returned status, current and projected rule values, coverage, data dates and mandate version. Preserve the check_id. Do not place an order or record a decision.
The amount is an illustration of how to state funding, not a suggested investment. Where an actual proposal names a security, give its ISIN or clarify the listing rather than letting the AI agent guess.
Read the result carefully. ok concerns the evaluated rules and evidence. breach identifies a rule conflict. cooling_off identifies a pause under your mandate. no_rule, unavailable and unclassified do not establish that the proposal is suitable or fully assessed. None predicts an investment return.
Completion check: find the saved check under AI agents → Activity and compare its reasons, dates and mandate version with the tool response. The saved verdict describes the evidence when that check ran. Calling check_action again creates a new check using then-available data.
When you have made your own decision, state it explicitly:
My decision for check_id [paste the returned ID] is deferred. Record that outcome with this note: “[my own reason].” Show the confirmed record and decision date. This statement is not an order instruction.
Supported outcomes are proceeded, abandoned and deferred. The outcome is recorded once. If the reply is interrupted, read that exact check back before attempting another write. The record documents what you reported; it does not prove a broker executed anything.
Step 5: Build your second brain for investing
Your investing second brain gives a future conversation something more reliable than a reconstructed memory of why you acted. The thesis capture and full journal-review workflow below requires Premium.
- Open a security's dossier in Research, then choose Record thesis.
- Write your own reason, a result that would change your view, and an optional expected observation. Keep captured Research facts separate from your interpretation.
- Choose Save as draft while the view is unfinished. A draft creates no review reminder.
- When ready, choose a future review date and Activate thesis. Activation records a version and its associated journal entry.
- Use View saved thesis version to revisit that version. Update thesis can create a revision while the earlier activated version remains preserved.
Ask a permitted connected AI agent to retrieve your current view:
Use get_own_record for ISIN [the instrument I specify]. Quote my statement and falsifier verbatim as my own words, with their recorded date, state and review date. If identity or access is unresolved, say so. Do not substitute your own thesis.
The tool returns a current projection, not an arbitrary historical version. Read an exact historical version in Investboard.
For a check that began with your AI agent, choose Adopt into the journal in Activity. Review the prefilled text, edit it into your own reasoning, optionally add a falsifier and future review date, then confirm Save entry using the displayed hold control. The entry retains its stored link to the originating check.
Completion check: open the saved entry and follow View original mandate check. Compare the exact check with the one you adopted. Matching dates or tickers alone do not establish a relationship.
Step 6: Complete a due review
When a review is due, open Reflection → Journal → Review past reasoning.
- Write what actually happened before revealing the original reasoning.
- Choose Lock observation when that observation is ready.
- Compare the revealed reasoning and falsifier with your observation. Write how sound the reasoning was, separately from the investment's outcome.
- Choose Confirmed, Refuted or Still open, then deliberately confirm Record review.
- Check the Review recorded receipt and open View saved entry and review.
When browser storage is available, an unfinished review is kept in that tab for up to 24 hours, with recovery wording after a reload. It is not saved to your account until confirmed. Keep the tab and account context intact while working.
A connected AI agent can use get_journal to read supported entries and returned verdicts. It cannot write the observation, assessment or review verdict. The journal tool also does not return the full observation and assessment; open the saved review in Investboard to read those words.
Still open means a review was recorded and remains undecided. A missing verdict means no review was returned. Neither means the thesis was confirmed.
When a result is missing or unclear
- No connection appears: complete sign-in, enable the connection in the conversation and request a policy read. Check the account and connection limit.
- A tool is denied: inspect the required scope and plan access. Reconnect with the permission you intend to grant; do not treat denied access as an empty record.
- No thesis is found: distinguish an instrument resolved without a thesis from an unresolved identity or ambiguous ticker. Clarify the ISIN or listing.
- A check is unavailable: inspect coverage, freshness and the returned recovery guidance. Repair the underlying data before requesting a new check.
- Activity looks incomplete: your AI agents' history covers their checks, not every check made inside Investboard. Use exact stored links to move between records.
- A save reply is uncertain: inspect the authoritative saved record before retrying. An item disappearing from a due queue alone does not prove a review was saved.
Keep control of access and shared information
Review and revoke external connections under AI agents → Connections. Revoking access stops future authorized reads and writes; it does not recall files or text already shared with another provider.
For a connection granted portfolio:read, the Portfolio analytics switch can disable those three read tools while leaving its other granted permissions available. Older connections need fresh OAuth approval for this scope; turning the switch on cannot grant access by itself.
The external AI agent's own processing, retention and training terms apply to information it receives. Investboard's built-in AI processing commitments describe that built-in feature, not every external AI agent. Leave unnecessary personal information out of free-text instructions and prompts.
For your next conversation, start with one task: read the rules, check the stated change, or retrieve your own recorded reasoning. Keep the tool response beside the explanation, and keep every decision in your own words.
Start your next AI conversation with your rules
Choose your AI agent, install the four skills and start with your own files. Add an authorized MCP connection for supported account data; portfolio analytics and saved history require Premium.
Choose your AI agent →Prefer to start with a file? Create and download your investing.md for free. No account needed.
Common questions
How do I start using AI in Investboard?
Record your mandate first and verify your imported data. Then open Investboard AI or export investing.md. For current policy reads and saved checks, connect a compatible AI agent through MCP and review its requested permissions.
Does uploading investing.md connect my account to ChatGPT or Claude?
No. The file contains your policy at its composition date and grants no account access. MCP requires separate sign-in and permissions. A file does not update itself or ensure that a model follows every rule.
Can AI trade for me or change my mandate?
Investboard's tools execute no orders and cannot change your mandate. check_action saves a check; record_decision records a decision you state. A separate broker connection in your AI agent has its own capabilities and permissions.
Which capabilities require Premium?
Active Free accounts support one MCP connection and the core policy, status, check and decision-recording tools. Premium adds unlimited connections, supported check-history, thesis and journal retrieval, and portfolio analytics. Each call also needs its matching granted permission.
Can my AI agent complete a journal review?
The MCP tools can read supported journal records, but cannot write a review. Record the observation, assessment and verdict in Investboard. An open verdict is a saved review that remains undecided; a missing verdict means no review was returned.
Apply this to your portfolio
Choose your AI agent and follow its installation steps for the four skills or an MCP connection. Start free with files you provide. Account access requires separate authorization; portfolio analytics and saved history require Premium.
Sources
- Investboard MCP: Werkzeuge und Berechtigungen · Investboard · Accessed: 2026-10-13
- investing.md: Format und kostenlose Arbeitsblätter · Investboard · Accessed: 2026-10-13
- Connect and test your plugin · OpenAI · Accessed: 2026-10-13
- Get started with custom connectors using remote MCP · Anthropic · Accessed: 2026-10-13
Related reading
Strategy & portfolio
Investboard in ChatGPT: a second brain for investing
Strategy & portfolio
AI investing needs a mandate, not a better prompt
Strategy & portfolio
Top 5 AI investing setups: strengths and weaknesses
Behavior & discipline
Investment Policy Statement for private investors: your mandate
Strategy & portfolio
Portfolio analysis with Investboard: A practical portfolio check