Agent Mode lets BidBot operate as an active bidding agent rather than a question-and-answer assistant. It can search several data sources at once, interpret dense tender documentation, perform structured analysis, and take actions on your behalf.

These can be used individually or combined in a single request.

Tooling is the set of modular capabilities that let BidBot perform specific actions inside BidScript. It is how BidBot grows alongside the platform — new capabilities appear without changing how you interact with it.

Your organisation's policies, templates and reference materials — both short-form facts and long-form narrative evidence.

BidBot searches the Bid Library by default. Every response includes a source trace showing which documents and sections it drew on, so you can verify any claim before it reaches a submission.

Three ways of asking, depending on what you need:

Always select your libraries first. See Using Specific Libraries.

Your archive of pre-approved question-and-answer pairs, mostly drawn from past bids.

Ask BidBot to search it when you want to reuse or build on an existing answer:

Specificity matters most here. Exact question text or distinctive keywords return the best matches, because the Q&A Store is searched on the words used rather than on meaning. See Smart Search.

Where the BidSearch add-on is enabled, BidBot can search live and historical tenders, contract notices and planning applications.

Narrow the search by document type, region or buyer, value range, publication dates, keywords, CPV codes or buyer type — and say whether you want a list or detailed summaries.

BidBot returns a structured list, each result with a summary and a link to the original source.

BidBot can search public sources when you explicitly enable it. It is never enabled automatically. See Web Search.

Agent Mode is not only about retrieval. BidBot can create and change things for you.

When you have developed a strong answer with BidBot, add it straight to your Q&A Store:

You can also paste a question and answer into the chat and ask BidBot to categorise and upload it. This is how good material gets captured at the moment it is written, rather than being lost in a chat nobody reopens.

If you notice something missing, out of date or wrong while working with BidBot, you can raise a flagged item without leaving the conversation.

BidBot creates the flag, which then appears on the Bid Library's Health dashboard, recording which library it relates to, who raised it, what needs adding, why it matters, and an example of the kind of document that would resolve it. From there it can be assigned and resolved.

This is deliberately low-friction: the person who notices a gap is rarely the person who will fill it, and asking them to go and find the Health dashboard is how gaps go unrecorded. See Bid Library health dashboard.

BidBot can generate a task pipeline from tender documents, then restructure it, move dates and assign people. See Building your project pipeline.

When you reference a project, BidBot can reach everything in it — uploaded tender documents, extracted questions, evaluation criteria, tasks and progress data.

Name the project explicitly:

This keeps BidBot focused on the relevant project rather than searching more widely than you intended.