Agent ecosystem opportunity report
188 scored opportunities for new agent tools, with the six component scores, confidence, and the strongest evidence for and against each of the top 10.
Executive summary
188 opportunities are scored: 108 integration gaps (measured demand) and 80 coverage gaps (hypotheses).
The highest-scoring integration gap (measured demand) is Better agent tooling for Snowflake at 67.4; the top 10 range from 61.8 to 67.4 on a 0–100 scale.
85 opportunities are rated Medium confidence and 103 Low; none is rated High.
536 of 1,283 evidence items (41.8%) are heuristics rather than measurements.
The distribution component is the same 79 for every opportunity, so it shifts all scores equally and does not change the ranking.
Every figure above is computed from the database DER
Methodology
How this report was produced, and its limits
Generated on 2026-11-21 from the AgentSidekick database. The latest completed ingestion step finished at 2026-11-21 13:23 UTC. Tables queried: opportunities, opportunity_evidence, market_segments, demand_signals, integrations, platform_details. The catalogue holds 73,820 active products, of which 9,469 meet the data-completeness bar and have a profile page.
Formula: overall = 0.25 × demand + 0.20 × gap + 0.20 × monetization + 0.15 × growth + 0.10 × feasibility + 0.10 × distribution, each component on 0–100 (see pipeline/intelligence.mjs). Growth = 50 + 50 × tanh(ln(segment launch ratio ÷ catalogue launch ratio)), using launches in the last 90 days vs the 90 before (each +1), and 50 when there are fewer than five dated launches. For coverage gaps, demand is the parent category's star percentile multiplied by a fixed discount (0.6) and labelled heuristic. Distribution = the share of tracked agent platforms that document MCP support. Confidence is Medium only when an integration has at least 50 issue mentions and 10 indexed products; everything else is Low. Opportunities were last scored at 2026-11-21 13:23 UTC.
Classification method: product type comes from the source (registry entries are MCP servers by definition) or is inferred from repository topics; categories, capabilities, integrations and industries come from keyword rules in the editable taxonomy, so they are derived, not checked by a person. Curated records were researched against the official sources listed on each record.
Known limits:
- Scores are heuristics. Monetization and feasibility are rules of thumb (business-focused integration, public API) with no revenue or engineering data behind them.
- Demand for integrations is the number of public GitHub issues matching "<name>" "mcp server". It measures developer discussion, not willingness to pay, and common-word names inflate it.
- Demand for capabilities is borrowed from the parent category's total GitHub stars; it was not measured for the capability.
- Gap counts products with a complete record and a repository pushed in the last 180 days. Products that offer a capability without naming it are not counted, so real gaps may be smaller.
- Categories, capabilities and integrations are assigned by keyword rules over names and descriptions (derived). Products that do not name a capability are missed, and generic words can produce false matches.
- Only 1 daily metrics snapshot(s) exist in product_metrics, so growth is measured from launch counts, not from changes in stars or downloads.
Note. Claim labels: source = stated by the publisher or registry; curated = checked by us against cited sources; derived = computed by a documented rule; heuristic = a rule of thumb with no measurement behind it. Full method: methodology.
How to read these scores
An opportunity score is a way to sort ideas for further research, not a forecast. Each of the six components is on a 0–100 scale and is either measured from the database (demand, gap, growth, distribution) or set by a rule of thumb (monetization, feasibility). The weighting favours demand and gap because those are the components with the most direct evidence.
Opportunities by confidenceHEU
Unit: opportunities. Top 2 shown.
Show the data as a table
| opportunities | |
|---|---|
| Low | 103 |
| Medium | 85 |
Note. Confidence is deliberately conservative: it rises to Medium only with both a measurable demand signal and an established product base. No opportunity can reach High without data the project does not yet collect (search volume, pricing, install counts).
Integration gaps (measured demand)
108 integration-gap opportunities are scored, one for every tracked integration with a demand measurement. Their average overall score is 52.0; the average demand component is 50.0 and the average gap component 40.0. Demand here is the count of public GitHub issues that mention the service together with "mcp server", turned into a percentile against the other integrations.
Top Integration gaps (measured demand) by overall scoreHEU
Component scores are 0–100. Overall is the weighted sum described in the methodology.
| Opportunity | Demand | Gap | Monet. | Growth | Feas. | Distr. | Overall | Confidence |
|---|---|---|---|---|---|---|---|---|
| Better agent tooling for Snowflake | 45 | 51 | 85 | 87 | 80 | 79 | 67.4 | Medium |
| Better agent tooling for Confluence | 61 | 55 | 65 | 72 | 80 | 79 | 66 | Medium |
| Better agent tooling for Home Assistant | 69 | 60 | 35 | 89 | 80 | 79 | 65.5 | Medium |
| Better agent tooling for Linear | 86 | 46 | 65 | 29 | 80 | 79 | 64 | Medium |
| Better agent tooling for Asana | 43 | 83 | 65 | 50 | 80 | 79 | 63.8 | Low |
| Better agent tooling for SQLite | 96 | 73 | 35 | 9 | 80 | 79 | 62.9 | Medium |
| Better agent tooling for Chroma | 52 | 94 | 35 | 50 | 80 | 79 | 62.2 | Low |
| Better agent tooling for HubSpot | 38 | 33 | 85 | 88 | 80 | 79 | 62.2 | Medium |
| Better agent tooling for Neon | 47 | 98 | 35 | 50 | 80 | 79 | 61.8 | Low |
| Better agent tooling for Sentry | 76 | 68 | 35 | 42 | 80 | 79 | 61.8 | Medium |
| Better agent tooling for BigQuery | 41 | 55 | 85 | 48 | 80 | 79 | 61.4 | Medium |
| Better agent tooling for Slack | 94 | 14 | 65 | 41 | 80 | 79 | 61.4 | Medium |
| Better agent tooling for Jira | 84 | 28 | 65 | 36 | 80 | 79 | 60.9 | Medium |
| Better agent tooling for Puppeteer | 50 | 89 | 35 | 50 | 80 | 79 | 60.7 | Low |
| Better agent tooling for Supabase | 89 | 16 | 35 | 78 | 80 | 79 | 60.1 | Medium |
Coverage gaps (hypotheses)
80 coverage-gap hypotheses are kept (the highest-scoring specific capabilities inside large parent categories). Their average overall score is 56.3 and their average demand component 39.8. Demand here is not measured: it is the parent category's GitHub-star percentile, discounted because interest in a parent category says little about one capability inside it. Treat these as questions to research, not ranked opportunities; every one is rated Low confidence and they should not be compared directly with the integration gaps above.
Top Coverage gaps (hypotheses) by overall scoreHEU
Component scores are 0–100. Overall is the weighted sum described in the methodology.
| Opportunity | Demand | Gap | Monet. | Growth | Feas. | Distr. | Overall | Confidence |
|---|---|---|---|---|---|---|---|---|
| Coverage gap: Interview Scheduling | 29 | 96 | 65 | 78 | 60 | 79 | 65.1 | Low |
| Coverage gap: Chat with Documents | 59 | 99 | 65 | 17 | 60 | 79 | 64 | Low |
| Coverage gap: Task Planning | 58 | 95 | 65 | 17 | 60 | 79 | 63 | Low |
| Coverage gap: CLI Coding Agents | 53 | 81 | 40 | 74 | 60 | 79 | 62.5 | Low |
| Coverage gap: Notebooks | 37 | 89 | 65 | 52 | 60 | 79 | 61.7 | Low |
| Coverage gap: User Preference Memory | 54 | 91 | 65 | 16 | 60 | 79 | 61 | Low |
| Coverage gap: Resume Screening | 29 | 75 | 65 | 78 | 60 | 79 | 60.9 | Low |
| Coverage gap: Entity Extraction | 51 | 95 | 65 | 10 | 60 | 79 | 60.2 | Low |
| Coverage gap: Reranking | 59 | 79 | 65 | 17 | 60 | 79 | 60 | Low |
| Coverage gap: Session Persistence | 54 | 86 | 65 | 16 | 60 | 79 | 60 | Low |
| Coverage gap: Employee Onboarding | 1 | 97 | 65 | 88 | 60 | 79 | 59.8 | Low |
| Coverage gap: TypeScript Agent Frameworks | 58 | 81 | 65 | 13 | 60 | 79 | 59.6 | Low |
| Coverage gap: No-code Agent Builders | 58 | 79 | 65 | 13 | 60 | 79 | 59.2 | Low |
| Coverage gap: Background Agents | 36 | 98 | 65 | 24 | 60 | 79 | 59.1 | Low |
| Coverage gap: Logging | 40 | 98 | 65 | 17 | 60 | 79 | 59.1 | Low |
The top 10, with evidence for and against
For each of the 10 highest-scoring integration gaps (the kind with measured demand), the table shows the strongest piece of evidence on each side: measured evidence is preferred over heuristics, and higher-weighted components over lower. 0 of the 10 have no recorded evidence against, which usually means a component simply was not assessed, not that there is no risk.
Top 10 integration gaps: strongest evidenceDER
Each evidence item is tagged measured or heuristic.
| Opportunity | Overall | Strongest evidence for | Strongest evidence against |
|---|---|---|---|
| Better agent tooling for Snowflake | 67.4 | gap (measured): 26 indexed products mention Snowflake; 20 of them have a complete record and a repository pushed in the last 180 days. | demand (measured): 558 public GitHub issues contain the phrase "Snowflake mcp" (checked 2026-11-21); that is the 45th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. |
| Better agent tooling for Confluence | 66 | demand (measured): 1225 public GitHub issues contain the phrase "Confluence mcp" (checked 2026-11-21); that is the 61th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. | gap (measured): One project holds 90% of the 24,968 GitHub stars in this segment, so there is an established leader to displace. |
| Better agent tooling for Home Assistant | 65.5 | demand (measured): 1733 public GitHub issues contain the phrase "Home Assistant mcp" (checked 2026-11-21); that is the 69th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. | monetization (heuristic): Home Assistant is not primarily a paid business product. This is a rule of thumb, not revenue data. |
| Better agent tooling for Linear | 64 | demand (measured): 4243 public GitHub issues contain the phrase "Linear mcp" (checked 2026-11-21); that is the 86th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. | gap (measured): 1 indexed product(s) are published under a namespace matching Linear itself (Linear). The gap score was reduced by 25 points for that. |
| Better agent tooling for Asana | 63.8 | gap (measured): 8 indexed products mention Asana; 6 of them have a complete record and a repository pushed in the last 180 days. | demand (measured): 539 public GitHub issues contain the phrase "Asana mcp" (checked 2026-11-21); that is the 43th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. |
| Better agent tooling for SQLite | 62.9 | demand (measured): 9661 public GitHub issues contain the phrase "SQLite mcp" (checked 2026-11-21); that is the 96th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. | gap (measured): One project holds 97% of the 1,239 GitHub stars in this segment, so there is an established leader to displace. |
| Better agent tooling for Chroma | 62.2 | demand (measured): 895 public GitHub issues contain the phrase "Chroma mcp" (checked 2026-11-21); that is the 52th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. | growth (measured): Too few dated launches to measure momentum (0 in the last 90 days, 0 in the 90 before). |
| Better agent tooling for HubSpot | 62.2 | gap (measured): 25 indexed products mention HubSpot; 16 of them have a complete record and a repository pushed in the last 180 days. | demand (measured): 454 public GitHub issues contain the phrase "HubSpot mcp" (checked 2026-11-21); that is the 38th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. |
| Better agent tooling for Neon | 61.8 | gap (measured): 1 indexed products mention Neon; 1 of them have a complete record and a repository pushed in the last 180 days. | demand (measured): 639 public GitHub issues contain the phrase "Neon mcp" (checked 2026-11-21); that is the 47th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. |
| Better agent tooling for Sentry | 61.8 | demand (measured): 2741 public GitHub issues contain the phrase "Sentry mcp" (checked 2026-11-21); that is the 76th percentile of the 108 integrations tracked. Issue mentions measure developer discussion, not buyers. | gap (measured): One project holds 80% of the 1,345 GitHub stars in this segment, so there is an established leader to displace. |
Top 5 coverage-gap hypotheses: strongest evidenceHEU
Demand for these is a discounted proxy, not a measurement.
| Hypothesis | Overall | Strongest evidence for | Strongest evidence against |
|---|---|---|---|
| Coverage gap: Interview Scheduling | 65.1 | gap (measured): 7 indexed products match this capability by keyword. | demand (heuristic): The parent category Recruiting holds 639 products with 216,457 combined GitHub stars (48th percentile of subcategories). Interest in the parent is only a proxy — demand for this exact capability was not measured — so the percentile is discounted by 40% to 29. |
| Coverage gap: Chat with Documents | 64 | gap (measured): 3 indexed products match this capability by keyword. | growth (measured): 394 launches in the parent category in the last 90 days versus 277 in the previous 90. |
| Coverage gap: Task Planning | 63 | gap (measured): 8 indexed products match this capability by keyword. | growth (measured): 173 launches in the parent category in the last 90 days versus 121 in the previous 90. |
| Coverage gap: CLI Coding Agents | 62.5 | gap (measured): 24 indexed products match this capability by keyword. | gap (heuristic): Classification is keyword-based: products that offer this without naming it are not counted, so the real gap may be smaller. |
| Coverage gap: Notebooks | 61.7 | gap (measured): 16 indexed products match this capability by keyword. | demand (heuristic): The parent category Data Analysis holds 255 products with 338,162 combined GitHub stars (61th percentile of subcategories). Interest in the parent is only a proxy — demand for this exact capability was not measured — so the percentile is discounted by 40% to 37. |
What this does not show: whether an official vendor tool is about to ship, whether existing products are good, or whether anyone will pay. Each opportunity page lists the existing products it was compared against and the risks recorded for it.
Recommendations
Editorial: the judgement of AgentSidekick, drawn from the figures above. Not a measurement, not investment advice.
- Editorial
start validation with the 85 Medium-confidence opportunities; the 103 Low-confidence ones rest mainly on proxies.
- Editorial
treat monetization and feasibility as placeholders. 41.8% of the evidence is heuristic, so a quick check of vendor pricing pages and API terms will change scores more than any reweighting.
- Editorial
for the top-ranked opportunity (Better agent tooling for Snowflake, 67.4), read the evidence against first; it names the assumption most likely to be wrong.
- Editorial
do not rank the 80 coverage-gap hypotheses against integration gaps; their demand is a discounted proxy, so a score of 65 there is not comparable with 67 for a measured integration gap.
Sources
- GitHub issue search counts: docs.github.com (external)checked
- Official MCP Registry: registry.modelcontextprotocol.io (external)checked
- GitHub repository metadata: docs.github.com (external)checked
- AgentSidekick methodology: agentsidekick.dev (external)checked
Data behind this report
- Scored opportunities with evidence: /api/dataset/opportunities.jsonJSON, regenerated each build
Cite this report
Agent ecosystem opportunity report. AgentSidekick, 2026-11-21. https://www.agentsidekick.dev/reports/opportunity-report/
Figures change when the database is refreshed; cite the date shown.