Changelog

September 23, 2026

Features

Decide what "relevant" means for you. Tender settings now let you weigh your own relevance criteria next to the built-in fit to your services, and set the high relevance and notification thresholds and your relevance bands. A live preview shows how many tenders a change would put in High relevance, and Save and recalculate applies it right away with visible progress. See Selection and scoring settings.
Find tenders outside your CPV codes. Keywords are matched by meaning in any language, so a Latvian keyword can find a German tender. Negative keywords move unwanted tenders to a new Excluded view, and you can limit selection to your countries of interest.
Prefer countries without excluding the rest. Add Preferred countries to your selection profile: tenders from other countries lose 15 points by default, so they reach high relevance only on a very close match. See Selection and scoring settings.
See whether you qualify. Each high-relevance tender gets an Eligibility score next to its Relevance: its qualification requirements are checked against your company profile and Library documents. Latvian companies can fill the profile from the Enterprise Register, and Check eligibility checks a tender on demand. See Eligibility check.
Shortlist and dismiss tenders. Star tenders to your Shortlist, dismiss the ones you won't bid on with a reason, undo either, or act on many tenders at once. The High relevance, Shortlist, All, Excluded and Dismissed views keep them apart. See Matched tenders.

Improvements

One tender per procedure. A procedure published in both IUB and TED now appears once, with links to every publication and a Change history of corrections, deadline extensions and results. See Tender page.
Deadline first. Lists sort by the nearest deadline and show the days left, a Closing soon filter narrows them to the next 7 days, and a tender with no deadline says Not stated.
How a tender was found. Each row says Found by a CPV code or a keyword, and after a recalculation you can open the tenders that are new with high relevance or no longer high relevance.
A richer daily email. The morning email now shows relevance, eligibility, the deadline and how each tender was found. See Daily email.
Tender decisions in the audit log. Shortlisting, dismissing and undoing are recorded under the new Tender selection category.
Claude Opus 5.5 is now the default analysis model.

Fixes

Changing your services no longer empties the matched list while tenders are rescored, and the scores are updated in place.
A tender is emailed only once, even when it is published again in another source or corrected.
Tenders without any deadline no longer stay open forever: they count as open for 50 days after publication.
Closed tenders are no longer included in the daily email.

September 3-23, 2026

Features

Plan the year's procurements. A new Planning module under Tools keeps your annual procurement plan: plan lines with their threshold band, requests from colleagues that you accept or decline, approval, amendments and publication, a quarter roadmap, and exports in the EIS and CVP IS layouts. Start a procurement or market research straight from a plan line, and run an AI check over the whole plan. Covers Latvian, Estonian and Lithuanian rules and the EU directive. See Procurement planning.
Design the procurement before drafting the RFP. Compose RFP now starts with Procurement design: the agent asks the key questions round by round, you answer, defer, or mark each one not applicable, and drafting starts only once the design is approved (in guided mode, with Approve design and draft). See Composing an RFP.
Request document changes. After a composed RFP draft is validated, describe a correction in your own words for one document or the whole package. The validation also compares the package with your original brief and shows its Review coverage. See Compose validation.
Search inside your Library. Library documents are indexed for content search, including spreadsheets, CSV files, and older Word files. Each entry shows whether its content is searchable, with Retry indexing when indexing fails. See Library.
An audit log for your organization. Owners and admins can see who did what and when — sign-ins, member changes, documents, runs, exports, and settings — filter it by user, category, and date, and export it to CSV. See Roles and permissions.

Improvements

Credits follow the model that did the work. A run is charged from its actual AI model usage, so the provider you choose affects the price, and the credit history shows the provider behind every charge. See Credits and pricing.
A clearer analysis run page. A four-step bar shows where the run is and, once it finishes, the verdict. A scorecard summarises requirements and findings, and each requirement lists the evidence behind it. See Understanding your results.
Bills of quantities compared row by row. When a tender includes a bill of quantities, the analysis compares the bid's priced estimate with it line by line.
VAT basis on market prices. Each price row shows whether VAT is included, excluded, or exempt only when a source states it, and shows the offer terms. The Excel export carries both, plus the gaps and unknowns. See Market research.
Screening shows how a company relates to the bid. A company found in a bid is labelled as a current bid participant or as related to a mentioned company, and an unclear relationship is raised for your review.
The verifier can raise a compliance verdict as well as lower it when the evidence supports it.

Fixes

Older .xls and .doc files are converted before they are read, so spreadsheet row labels and document text are no longer lost.

September 1-2, 2026

Features

See what a procurement is actually buying. A new Items tab extracts every product and service line out of the technical specification, with quantity, unit, CPV code, lot, and source reference. Correct what the extraction got wrong, dismiss what does not belong, add what it missed, and approve the rest. See Items.
Find suppliers and prices for those items. Once items are approved, market research runs from inside the procurement: candidate suppliers with the comparable public contracts they have already won, one sourced price row per item with its basis, spec match, and date, and a written report, on the new Suppliers, Prices, and Report tabs. Export the price table to Excel. See Market research.
A supplier register. Every bidder you add becomes a supplier record under Tools → Suppliers: link it to its business-register entry, see the screening results from past analyses and every bid it has made, and merge duplicates. The bid form now offers existing suppliers as you type. See Supplier register.
Academy: OECD public procurement standards. A second quiz on the Tendergate Academy, alongside the Latvian Public Procurement Law: 30 questions on the OECD Recommendation on Public Procurement, in English, with every answer explained and cited to the source text. Open it from Academy in the top bar or at tendergate.ai/academy/oecd-public-procurement.

Improvements

One workspace for composing and analysing. Creating a procurement no longer asks you to choose between analysing bids and drafting an RFP. Compose is a tab inside the RFP group, and a composed RFP is locked with Finalise RFP before requirements, bids, and analysis follow. The RFP tabs are renamed Documents and Requirements, and evaluation criteria and lots now sit before the quality and integrity checks. See Setting up a procurement and Composing an RFP.
See the scope before the credits. Starting market research from Items shows how many approved items and distinct products the run will price before you confirm.
The activity feed narrates the pricing stage of a market research run, batch by batch.

Fixes

The prices Excel export downloads instead of opening inside the page.
Item extraction asks for confirmation before it starts, and the Items tab refreshes itself when the extraction finishes.

August 31, 2026

Features

The Assistant reads your scans. Once an analysis has transcribed a scanned page or image on a procurement, the Assistant can answer from it, reusing that text at no extra cost instead of reporting that it cannot read the file. See Assistant.
See the text taken out of a scan. Each document now carries the transcription a vision model produced, shown under Extracted text, and the document list badges whether text was read, nothing legible was found, or the read failed. The text is stored and reused, so later runs no longer re-read — or re-charge for — the same scans. See Working with documents.

Fixes

A document saved without a file no longer leaves Run AI analysis permanently disabled with no job running.
RTF, ODT, and ODP files upload with their contents instead of being stored as an empty reference, so their text reaches both analysis and the Assistant.
Assistant citation numbers now come from the search itself, so a citation always points at the document the claim actually came from.
The Assistant searches for something in the current question before saying it is not in the documents, rather than concluding that from earlier context.

August 6-26, 2026

Features

A rebuilt Assistant. Now named Assistant rather than Chat, it works your documents directly — searching for wording, opening the file, reading around the match — instead of querying a fixed index. Start a session to prepare your documents, and every claim comes back with a numbered citation to its source. A session left idle ends cleanly and restarts without losing the conversation. See Assistant.
Register screening in your exports. Analysis and comparison exports now carry an Eligibility and capacity sheet listing every register check per screened subject, with its status, how the subject was matched, the source register, and the date the data was current. See Comparing bids.
See which documents a verdict covered. Open a requirement in the comparison view and a Documents checked panel shows how many of the expected documents were reached, and which failed or were never reached.

Improvements

Subcontractors are found in the bid text itself. Register screening no longer depends on the analysis noticing a company: the proposal is swept for registration numbers, and every one that resolves in the business register is screened. A subcontractor named deep in an annex is now checked. See Understanding your results.
Register facts short of an exclusion raise a finding too, weighted by the kind of tender — a builder struck from the construction register is a major issue on a works contract and a minor note on a stationery contract.
Contradicted strengths are withdrawn. A strength praising a declaration that screening later contradicts is removed instead of standing beside the finding that contradicts it.
Watch the image reading happen. The activity feed now reports the pass that reads scans and images, with progress and a note when the per-run limit is reached, instead of looking frozen for long stretches.
Evaluation criterion names are short titles again, rather than internal keys or whole paragraphs.

Fixes

A tender uploaded only as a spreadsheet is detected in its own language. It previously fell back to English silently, so a Latvian tender came back reviewed in English.
Uploading the same file twice is refused and marked already uploaded in the uploader, instead of quietly doubling your document set.
Dragging several files at once uploads all of them.
The Peer Insights tab compares you only against tenders from a country the corpus covers, so an Estonian or Lithuanian procurement is no longer benchmarked against Latvian ones.

August 4, 2026

Features

Tendergate in German. Switch the interface to German with the language selector in the top bar. Analysis reports, research runs, and preparation checklists can be produced in German too. This documentation remains available in English and Latvian.
Verification you can trust. When the verifier cannot re-read the evidence behind an item, it is now marked Not verified instead of being confirmed by default. Evaluation criterion values and requirement-compliance rows are verified alongside findings, and the Excel export carries the verification outcome, note, and original status. Findings raised by register screening are labelled as register results rather than document-verified findings. See Understanding your results.

Improvements

Tender match reasoning in your language. The explanation of why a tender matched your profile is now written in your organization's interface language instead of always in English.
Funding research can target a specific country and filter sources by how recent they are, matching the options market research already had.

Fixes

The opportunities comparison table in funding research renders its rows again.
Price tables in PDFs with merged cells are read row by row, so a group subtotal is no longer mistaken for a line item's own total.

July 26-28, 2026

Improvements

Re-run a failed research run. A funding research run that failed can be started again from the failure panel, without leaving the page.

Fixes

Register screening no longer reports exclusion grounds as clear when the bidder could not be resolved in the Latvian register. Those checks now read Unknown, the card says the registration number was not found here, and the manual-check reminder still appears. Tax debt lookups also stopped failing on days when the State Revenue Service could not serve the requested date.
OpenDocument text and spreadsheet files (.odt, .ods) are parsed correctly.
The recommendation on the funding research overview is no longer cut off mid-sentence.

July 23, 2026

Features

See how peers structured similar tenders. When you run an RFP quality check, a new Peer Insights tab shows how other organizations that published similar tenders set up their evaluation, requirements, and documents — compared against your own RFP. Each insight links to the source tender, flags how likely it is to apply to you, and can be dismissed. Peer Insights are advisory: they never change your quality recommendation. See RFP quality review.

July 22, 2026

Features

Automatic bidder screening against public registers. For Latvian procurements, a new Eligibility & Capacity tab automatically checks the bidder, its subcontractors, and its owners and board members against open-data registers — sanctions, insolvency, suspended activity, VAT status, and tax debt — and raises a finding when an exclusion ground applies. Financial turnover, construction qualification class, and certified specialists are matched to the bidder and shown alongside. See Understanding your results.
Full company and person details, traceable to the source. Click any screened company or person to open its complete register record — status, address, beneficial owners, shareholders, and board members — with a link straight to the original data.gov.lv dataset behind every data point.

July 21, 2026

Improvements

Several model providers, one per run. Your organization can now save more than one AI provider profile under Account settings → Other model providers, each with its own OpenRouter API key and a model preset for the main, mid, and fast tiers. Saved profiles appear in the provider menu next to Anthropic, OpenAI, and Google, so you can keep one profile tuned for cost and another for quality and pick between them. The single OpenRouter key that used to live on the Billing details page has moved here and became your first profile. Every run records the profile it started with, so editing a profile later doesn't rewrite the history of runs that already used it.

July 17, 2026

Features

Start an RFP quality check with an upload. Quality checks now have a focused workspace under Tools: choose the full RFP package directly, follow its progress, return to previous checks, and export the results to Excel. A new account includes one complete quality check with no credit card required. See RFP quality review.
More national jurisdictions for RFP quality review. The Legal jurisdiction field now covers Estonia, Lithuania, Poland, and Germany in addition to Latvia and the EU directive. Auto-detect picks the right one from your documents, so Legal findings cite the national procurement law that actually applies. See RFP quality review.
Continue with Google. You can now create an account or sign in with Google from the authentication pages.

Improvements

Relevant Latvian complaint precedents in legal findings. For Latvian quality checks, Tendergate can search published IUB complaint decisions and cite a decision when it directly supports the issue found in your RFP.
Faster scanning across countries. The tender catalogue now shows each notice's country with a flag, and the All tenders and Matched tenders tabs use the same filters.
Quality-check categories, live activity, results, and Excel exports now stay in the selected language.

Fixes

Quality-check starts and reruns now validate access and available credits consistently, preventing confusing partial starts.
Registration, confirmation emails, and the destination after sign-in now preserve the language and product flow you selected.

July 12, 2026

Features

Discover tenders from 11 new national sources. You can now find opportunities from official sources in Poland, Lithuania, France, Austria, Czechia, Estonia, Norway, Portugal, Italy, the Netherlands, and Ireland. Newly imported open tenders from every source automatically enter the daily matching flow for your company profile. See Tenders.

July 1–11, 2026

Features

Read images, charts, and scanned documents. Analysis and RFP agents now inspect the actual content of diagrams, scanned pages, stamps, and signatures with a vision model instead of guessing from a filename.
Analyze in the procurement's language. Tendergate detects the RFP language and produces findings, summaries, and Excel exports in that language. You can still override it manually.
Open requirement and criterion details without losing your place. Click a row in bid comparison to see the offered value, verification reasoning, and every evidence quote in a focused dialog.

Improvements

Search the CPV browser by code or description when setting up tender preferences.
Organization owners and admins can resend pending invitations and see when each invitation expires.

Fixes

Non-English analyses and spreadsheet exports no longer fall back to English for statuses and structured values.

June 30, 2026

Features

We're now Tendergate. MitigateTenders has become Tendergate, with a new home at tendergate.ai. Existing procurements.mitigate.dev links redirect automatically, so nothing you've bookmarked breaks.
Choose the legal jurisdiction for RFP quality review. When setting up a procurement you can now pick which procurement-law rulebooks the quality check uses for its Legal findings — auto-detect, the harmonized EU directive, or EU plus Latvian national law. See RFP quality review.

Improvements

Two composition modes instead of three. Drafting an RFP or a bid now offers just Autopilot and Guided. Guided is adaptive — it asks for the issuer or vendor details it needs only when the plan actually calls for them, folding in what used to be a separate Interview mode. See Composing a bid and Composing an RFP.
Faster AI runs. Multi-step agent runs now reuse cached conversation context, so long analyses and compositions respond quicker and use fewer tokens.

Fixes

Paused bid composition runs now finish cleanly when resumed, and the plan only appears once composition has completed.
Sign-in now returns you to the page you were heading to instead of a generic landing spot.
Matched-tender period stats are now based on each notice's publication date, so the counts match the period you picked.
German tender notices now link back to their original source, and missing country filter labels in the tenders catalogue have been restored.

June 19, 2026

Features

Invite your team. Organization owners and admins can now invite colleagues by email from account settings, assign roles (owner, admin, member), change them later, and remove people who leave — invitees get a branded email and set their own password. See Roles and permissions.
Bring your own AI provider. Connect your organization's own OpenRouter API key in account settings and pick the model for each tier, so analysis and every other AI run uses your account instead of ours.
Guided product walkthrough. A new Showcase in the top navigation walks you through the platform chapter by chapter — setting up a procurement, running an analysis, comparing bids, and composing a bid — in English and Latvian.
Daily tender match alerts. Owners and admins now get a daily email listing new tenders that match your organization above 80% confidence, in your preferred language, so strong matches reach you without checking the catalogue. See Tenders.
Redeem a promo code at sign-up. The registration form now has an optional promo code field, so you can apply a code when you create your account.

Improvements

RFP composition now mirrors the bid flow. Drafting an RFP plans inside the workspace instead of a separate setup interview, asks for the specific facts it needs through short structured forms rather than open questions, lets you adjust the planned documents, and adds step-by-step navigation. See Composing an RFP.
Requirement compliance in your exports. Excel exports now include requirement compliance — a per-requirement status sheet for a single run, and a requirements-by-vendor matrix in the all-bids comparison export.
Quality review findings as a scannable table. RFP quality review now lists findings in a table with severity, category, finding, and source columns, so you can scan everything at a glance and expand any row for the full details and references.

Fixes

Answering a clarification during analysis no longer breaks the run when it resumes.
The composition quality check no longer hangs indefinitely if processing stalls.
Funding research verifies its sources more strictly, so results no longer cite weak or unreachable references.

June 15, 2026

Features

RFP requirements. Extract a structured list of every requirement in your RFP, review and approve them in one matrix, and have them feed your preparation checklist, evaluation criteria, and lots automatically — so everything stays anchored to what the RFP actually asks for. See RFP requirements.
Requirements matrix in bid comparison. Compare every approved RFP requirement against each vendor side by side, with a per-requirement compliance status and evidence on demand. See Comparing bids.
Requirements on analysis results. Each analysis run now has a Requirements tab showing every requirement's status (met, partial, failed, unverified, or not applicable), the value the vendor offered, and the exact RFP clause it came from.
Smarter bid drafting. Bid composition is now driven by the RFP's requirements: the Documents tab flags any mandatory requirement that isn't yet covered, and you can download the whole bid as a ZIP package with a readiness summary. The agent also asks for the specific vendor facts it needs through short structured forms instead of open-ended questions. See Composing a bid.
Chat over your analysis results. The procurement assistant can now answer from completed bid analyses — gaps, compliance against criteria, lot participation, and the recommendation — and cite the analysis, not just the source documents. See Document chat.

Improvements

Live preparation status while an analysis is starting, so you can see it extracting requirements and detecting the procurement domain before specialists begin.
Apply only the validator findings you select, or re-run validation against the current draft without re-drafting everything.
The preparation checklist and evaluation criteria now stay in sync with your approved requirements — re-extract and re-approve to update them.

Fixes

Market research no longer cites snippet-only sources from blocked or low-quality domains.
Funding research runs can now be deleted.
Clearer empty states across procurements, bid comparison, and research so it's obvious what to do next.

May 26, 2026

Features

RFP composition, rebuilt. Drafting an RFP now uses the same flow as bids: a short setup interview produces a plan you approve, the agent writes the documents as files you download (individually or as a ZIP), and an automatic review always pauses for you before anything is final. See Composing an RFP.
Export a single analysis run to Excel. Download one run's full results as a spreadsheet, in addition to the all-bids comparison export.
Forgot your password? Reset it yourself from the sign-in page — no need to contact support.

Improvements

Claude Opus 4.8 is now the default analysis model.
7z archives are accepted and unpacked automatically, and signed document bundles (EDOC) with nested containers are unpacked all the way through.
Tenders now include Lithuanian notices, and your tender matches re-score automatically when you change your services or CPV preferences — no waiting for the next daily run.
Saved filter presets now work on the Matched tenders tab too.

Fixes

Document row menus stay visible and usable; the experimental mind map view was retired.

May 10, 2026

Documentation

New compose section pages. Documentation now covers the full bid composition flow end to end: the setup interview (where the agent drafts a plan you approve), composing a bid itself, and compose validation — the two-pass automatic review (coverage + content) that runs after composition completes.
Sandbox architecture, in plain English. How AI agents work and document parsing have been rewritten to explain the per-run sandbox isolation, on-demand semantic search, and the actual agent roles in use today.
Glossary expanded. Added 18 new entries covering compose plans and artifacts, sandbox isolation, autopilot/guided/interview modes, validation findings, library auto-pull, and more.
Pricing and security updates. Credits and pricing now lists every billable run type and typical cost ranges, including bid composition and revision rounds. Security and privacy explains the per-run sandbox isolation that keeps each run's documents separate.

April 20, 2026

Features

RFP integrity check — scan any uploaded RFP for signs of directed procurement: overly specific requirements, unnecessary certifications, tight timelines, narrow experience thresholds, vendor-specific language, and other fairness risks. Each finding quotes the RFP text that triggered it, points back to the section, and is categorised by severity. Run it before publishing your own RFP or when deciding whether to bid on someone else's. Available as a new Integrity check step in the procurement sidebar. Read more.

Improvements

Findings on the RFP integrity and quality reviews are now collapsed to one-line rows by default. Click any row to expand and see the full description, recommendation, and evidence — making long finding lists easier to scan at a glance.

Fixes

Market research and funding research runs now correctly charge credits on completion. Earlier runs completed without the credit deduction being recorded.

April 18, 2026

Features

Library document preview — preview library documents inline, download the original file, and expand or collapse the parsed content, without leaving the library view.
Inline edit for tender services and products — edit services and products directly on the tender detail page instead of opening a separate modal for each change.
Finish composing and go to analysis — hand a composed bid straight to the analysis agent with one click from the compose view.

Improvements

Tenders catalogue opens on your matched tenders by default and automatically re-applies the filter you used last time, so you return straight to the list you care about
Tender notice type and procedure type show human-readable labels instead of raw procurement codes
A read-only banner is shown while the compose agent is running, so you don't edit content that will be overwritten when the agent finishes
RFP uploads and Library uploads are now visually distinct so it's clear where each file is going
Disabled "Generate checklist" button explains which prerequisite is missing instead of just being greyed out
Analysis verdict shows a stat grid and severity distribution so you can read the outcome at a glance without scrolling through the full findings list
New Verification coverage tab on analysis runs lists every RFP clause, numerical claim, signature, attachment reference, and entity mention the AI verified — including the ones that passed — so you can see the full picture of what was checked, not just what failed
Latvian documentation now uses "nolikums" in place of "RFP" throughout

Fixes

Guided bid composition correctly accepts your clarification reply instead of re-asking the same question
Analysis surfaces sub-agent errors and double-checks every pass/fail criterion, so compliance gaps no longer slip through silently
Bid rows show pending, paused, and failed states, and starting a duplicate analysis run is blocked while one is already active
Large xlsx comparisons no longer overflow the sub-agent context window mid-run
Opening a procurement no longer errors when an RFP document is still parsing

April 17, 2026

Features

Signature detection — uploaded documents are automatically scanned for digital signatures. eDoc / ASiC-E containers are cryptographically verified, and PDFs are checked for embedded signatures. Document rows show a signature badge with the signer's name and signing time. Signatures on an eDoc container automatically apply to every document inside.
Signature-aware analysis — AI agents now know which documents are signed and by whom, so criteria like "form signed by authorised representative" are verified directly from the signature instead of being inferred from the document text.
Submittable documents — the Preparation Checklist now lists each specific document the vendor has to submit (application form, bank guarantee, ISO certificate, etc.) as its own row with a dedicated "Attach" action.
Library auto-match — existing documents in your Organization Library are automatically matched against each submittable. A valid ISO 9001 certificate already in the library is offered as a one-click attachment instead of asking you to upload it again.
Library document classification — library documents are automatically classified after parsing (certificate, financial statement, reference letter, template, etc.). The classification is visible in the library preview and is used by the auto-matcher.

Improvements

Document upload zone redesigned with explicit "Choose files" and "Choose folder" buttons
Procurement overview redesigned as a guided setup checklist so you can see at a glance which steps are still pending
Document detail page gained a Signature info card listing every signer, signing time, and signature format
IUB and EIS source links added to tender detail pages so you can jump to the original notice in one click
Empty states across the app use positive copy with next-step guidance and documentation links
Library tab shows per-category counts and supports full-text search

Fixes

Logging in now lands directly on the procurements list instead of a separate console view
"Extract from RFP" buttons show a loading indicator while running
Undo history is cleared after the editor loads AI-generated content, so Ctrl+Z no longer wipes the draft
Checklist preserves scroll position when you toggle items
Editor no longer errors when navigating between procurements
Latvian translations fixed for authentication messages and agent event broadcasts

April 8, 2026

Features

Updates page — public-facing changelog page accessible from the landing navigation and footer. Curated timeline layout showing platform updates in all 5 languages (EN, LV, LT, DE, EE).
Advanced tender filters — the Matched tenders tab now supports filtering by region, CPV code, notice type, and full-text search. Filter buttons show per-tier match counts.
Real-time status updates — analysis bid rows and procurement cards now broadcast live status changes, so you see progress without refreshing the page.

Improvements

Processing badge shown while bid documents are parsing or indexing
All index pages standardized to consistent max-w-7xl content width
RFP upload and bid pages are now fully scrollable instead of only the document list
Criteria extraction blocked when criteria already exist, preventing duplicates
30-criteria soft cap enforced to keep analysis focused
Users are soft-deleted instead of hard-deleted, preserving audit history

Fixes

Compose editor now unlocks after AI run completes (PROC-26)
Tender detail pages no longer break when notices have a source field (PROC-19)
Global stats hidden when tender filters are active to avoid misleading counts (PROC-20)
All hardcoded English strings translated to support Latvian locale (PROC-22, PROC-23)
AgentSet namespace and document names use IDs to avoid 64-character limit
3 production job failures resolved from discarded jobs audit
Tender matching UX improvements and legal locale fallback

April 4, 2026

Features

Console redesign — Mission Control — the console dashboard has been completely redesigned as a full-width mission control layout. Four action buttons (Analyze Bids, Draft RFP, Market Research, Funding Research), a unified activity feed showing all recent work across every agent type, a matched tenders panel with match scores, and platform-wide metrics (active tenders, matched, running agents, credits).
Tenders catalogue — browse and filter all active Latvian public procurement tenders. Two tabs: "All tenders" with full-text search, CPV code filtering, and saved filters; "Matched for you" showing tenders matched to your organization's profile with confidence scores. Includes tender detail pages with match analysis and deadline tracking. See Tenders.
CPV code browser — hierarchical tree modal for browsing and selecting Common Procurement Vocabulary codes when filtering tenders.
Billing Portal — access invoices, payment methods, and billing history directly from the user dropdown menu. Invoices include VAT and company details.
Checklist-driven composition — the bid composition agent now uses your prep checklist as a structured guide. Each checklist item is tracked during composition, and the agent must address all items before finalizing. Start a composition directly from the checklist page with one click.
Library document management — save and remove documents to/from the organization library directly from document rows.
12 new domain knowledge modules — expanded AI domain expertise covering additional procurement sectors.

Improvements

"Procurements" renamed to "Analysis" in navigation, with Tenders added as a new top-level nav item
Buy button added to credit history modal for quicker credit purchases
Invoices now include VAT and company details
Notifications moved to bottom-left corner with slide-in animation
Tender settings page with company details and service category management
Document row actions redesigned with three-dots dropdown menu
Checklist status indicators show which items the agent has covered

Fixes

Extensive Latvian translation improvements across the platform
Clarification response and checklist navigation issues resolved

March 26, 2026

Features

Blog post 13 — Latvia procurement reform 2026, covering the latest regulatory changes. Available in all 5 blog languages.

Improvements

Improved page load performance with compressed images and lazy loading
Better social media sharing with Open Graph tags and default preview image

March 24, 2026

Features

Funding Research — new AI research agent that discovers grants, subsidies, EU programmes, and other funding opportunities relevant to your procurement or business needs. Upload context documents, describe what you're looking for, and the agent searches structured databases and the web. Results include categorized opportunities with eligibility requirements, deadlines, funding amounts, and application links. Includes XLSX export for offline review.
Scope Scout — the analysis agent now detects which lots each vendor is actually bidding on before starting the main analysis. This prevents false positives on lots where the vendor didn't submit a proposal, and focuses the analysis on relevant lots only.

Improvements

Mobile responsiveness improvements across landing pages, footer, authentication screens, and navigation menus
Dedicated views for Quality Check and Prep Checklist results (previously inline)
Lots tab added to Compare & Export as a proper route with "not participating" status display
Market research findings redesigned with comparable/insight type classification
Source URL validation added to analysis findings for more reliable provenance links
Admin token analytics expanded with market research and funding research tracking

Fixes

Sidebar now shows correct status for failed analysis runs
Queued bids show "Queued" instead of "Analyzing" in procurement overview
Verifying status displays correctly across sidebar and bid cards
Lots matrix table styling aligned with criteria table
Mobile sidebar backgrounds render as solid (no transparency bleed)
XLSX export moved to background processing to prevent timeout on large reports
Discarded bids no longer conflict with vendor name uniqueness checks

March 19, 2026

Features

Market Research — AI-powered market research agent that searches the web, reads product pages, and compiles structured findings with verified sources. Describe what you need, optionally upload reference documents, and the agent produces categorized findings (products, suppliers, regulations, risks, recommendations) with confidence scores and clickable source URLs. Two modes: Autopilot and Guided. See Market Research.

Improvements

Redesigned findings UI with a Products/Services table and a separate Insights view
Finding cards now show full descriptions on expand, clickable source URLs, and confidence rationale
Finding descriptions and executive summaries render as formatted markdown
Language selector added to market research form

March 18, 2026

Features

Self-service registration — sign up directly from the landing page with email verification. Registration collects name, email, organization name, country, and an optional referral source.
Blog — 12 procurement-focused articles covering AI in procurement, bid rigging detection, EU regulatory changes, and common procurement pitfalls. Available in English, Latvian, German, Estonian, and Lithuanian.

Improvements

Excel export for RFP quality review findings with severity-coded formatting
Continue button for failed bid composition runs — retry without starting over
RFP quality review now recognizes procurement templates and tailors analysis accordingly
Organization info section added to account settings
Improved welcome banner design on the console dashboard

March 16, 2026

Features

Console dashboard — new landing page after login with action cards, recent activity feed, and status overview with procurement and bid counts. See Setting up a procurement.

Improvements

Deleting a procurement now archives it instead of permanently removing it — credit history and usage data are preserved
Quality Check and Prep Checklist status indicators update in real-time during agent execution
Responsive hamburger menu for mobile navigation on the landing page

Fixes

Missing translations on bid pages (vendor name and received date labels)

March 13, 2026

Features

5 domain knowledge modules — specialized AI knowledge for construction materials, food/catering, IT/digital services, medical/pharmaceutical, and technical equipment procurements. Each includes classification systems, regulatory references, and common compliance mistakes. See Domain knowledge.
Qualification matching engine — universal module that standardizes vendor qualification validation with structured matching modes and 3-source cross-validation.

Improvements

Domain knowledge now applies across all AI features — analysis, verification, criteria extraction, and RFP composition
Analysis run history — switch between multiple analysis runs per bid using the run selector dropdown
Redesigned RFP review finding cards with sectioned layout (issue, recommendation, evidence)

Fixes

Procurement status incorrectly showing "Analyzing" when a bid had multiple completed runs
Untranslated bid header text

March 12, 2026

Features

Compose RFP — AI agent drafts complete procurement documents from scratch in three modes: Autopilot, Guided, and Interview. See Composing an RFP.
Organization Library — store reference documents, templates, and standard clauses that the compose agent searches automatically when drafting.

Improvements

Rich text editor with full toolbar for composed documents (headings, lists, tables, bold/italic, links, alignment)
Floating table toolbar for quick row/column editing
DOCX export for composed documents with formatting preserved
Placeholder highlights (yellow) mark sections needing your input, with badge counts in sidebar
Auto-save with visual indicator in editor toolbar
Run Quality Check on composed RFPs
RFP documents automatically added to library when indexed
Redesigned procurement summary as a bid-centric dashboard
Real-time agent activity feed during composition
Clarification flow — agent can pause and ask questions, then resume

March 11, 2026

Improvements

Complete UI redesign with guided workflow navigation — step-by-step process instead of a file explorer
Persistent top bar with organization, credits, language, and profile controls
Procurement list page with status cards
Workflow sidebar with guided steps, bid list, and suggested next actions
Documentation site integrated directly into the platform at /docs

Fixes

Lots and criteria editing for the new layout
Checklist toggle causing a blank page
Language toggle button spacing

March 5, 2026

Improvements

AI-generated summaries render as formatted markdown in RFP review and checklist views
Run metadata (model, start time) shown in analysis, RFP review, and checklist views
Credit charging added for checklist generation
Agent Flight Recorder added to checklists — watch the AI work in real-time
Reduced token usage by not embedding full documents in RFP review and checklist prompts

March 4, 2026

Features

Agent Flight Recorder for RFP review — two-panel layout showing results and live agent activity

Improvements

Unified progress tracking across all agent-powered features

On this page