Glama.ai Emerging

Name: Agent Native Offer Audit — Glama.ai
Item: Glama.ai
Rating: 66
Author: Agent Native Offers

AUDIE Score: 66/100 · Audited 2026-04-13 · Website: https://glama.ai · Machine-readable: JSON

Pillar Scores

P1 Signal Architecture — 22/25

P2 Clarity Stack — 16/25

P3 Trust Envelope — 13/20

P4 Velocity Triggers — 7/10

P5 Gravity Design — 8/20

Executive Summary

Glama.ai is the highest-scoring company in the Emerging tier — at 66/100, it sits just above the Human-Dependent boundary and is genuinely on the agent-native path. Its Signal Architecture is near-perfect: llms.txt confirmed, OpenAPI spec live, rich schema markup, and the platform itself is MCP infrastructure. The critical drag comes from Gravity Design (8/20), where Glama captures vast usage data but exposes none of it back to agents — no memory API, no renewal signals, no compounding value indicators. The highest-priority fix is a documented fallback policy for the ~4% downtime gap, which costs nothing and immediately addresses the most glaring trust hole. With three targeted changes (fallback policy, API versioning, programmatic status endpoint), Glama could credibly reach 75+ and become a reference-grade Agent-Ready platform.

Strongest Signals

Best-in-class Signal Architecture llms.txt confirmed, OpenAPI spec present, rich schema.org markup on all key pages, and the platform IS the MCP infrastructure — agents can discover and access 21k+ tools through the gateway API. This is genuinely agent-native infrastructure.
Real Operational Transparency Public status page with timestamped incidents and 96.08% uptime figure. Per-tool access control and full call logging give agents meaningful control over what runs and a complete record of what did.
Precise, Parseable Pricing Three self-serve tiers with exact numbers for every limit and overage. No "contact sales" gatekeeping until enterprise scale. An agent evaluating whether to use Glama can get all cost information in one page visit.

Critical Gaps

No Substitution or Fallback Rules (P2-C: 1/5) With 96.08% uptime (meaning ~58 hours of downtime per 60 days), there is no documented fallback path. For agents running in production, the lack of degradation protocols or alternative routing is a significant trust gap.
No Behavioral Versioning (P3-D: 2/5) No API versioning strategy, no documented notice period for breaking changes, no version-controlled terms. Agents relying on the gateway have no programmatic way to know if the interface will change.
No Gravity Mechanisms for Agents (P5-B/C/D: 5/20) The platform captures immense usage data but exposes none of it back to agents. No memory API, no renewal signals, no compounding value indicators. Agents treat each session as fresh with no accumulated context.
Uptime Below Agent-Grade Threshold 96.08% over 60 days (~58 downtime hours) is below what production agent workflows typically require. No SLA or compensation policy found.

Priority Actions

All 20 Criteria

P1-A Structured Data — 5/5
Rich Schema.org JSON-LD confirmed on MCP servers page: Organization, BreadcrumbList, SearchResultsPage, and SoftwareApplication schemas for every indexed server (21,337 entries). Each entry includes author, license, OS, description, and GitHub URL.

P1-B Machine-Readable Pricing — 3/5
Three tiers ($9/$26/$80/mo) clearly structured in HTML with exact limits and overage costs. Not encoded in schema.org/Offer format, but granular and parseable.

P1-C llms.txt / Agent Layer — 5/5
Confirmed /llms.txt present at glama.ai/llms.txt with structured sections covering Products, User/Support resources, and Gateway documentation — explicitly formatted for LLM consumption.

P1-D API / MCP Availability — 5/5
OpenAPI spec at gateway.glama.ai/openapi.json confirmed. OpenAI-compatible API gateway with API key management. The platform IS the MCP infrastructure — 21,337 servers accessible via programmatic gateway.

P1-E Discoverability (GEO) — 4/5
Sitemap at glama.ai/sitemap.xml, rich structured data throughout, curated categories (86 total), described as "#1 Platform for Discovering MCP Servers." Strong AI retrieval optimization but no explicit GEO documentation.

P2-A Offer Completeness — 4/5
What (MCP hosting/gateway/registry), who (agents, developers, teams), how much (three tiers with exact $/month clearly listed on /pricing). Nearly machine-parseable from a single page.

P2-B Scope & Limits — 4/5
Explicitly stated: 3/10/30 fast hosted MCP servers per tier; $4/$3/$2 per additional server; 100k logs/month included; overage at $9/$6/$3 per 100k; log retention 30/90/180 days. All quantified.

P2-C Substitution Rules — 1/5
No documentation found on what happens during service degradation, server unavailability, or fallback routing. Status page shows 96.08% uptime but no SLA or fallback rules.

P2-D Conditional Logic — 3/5
Most conditions visible on pricing page (overages, tier gates). Cancel-anytime policy stated. Some features require contacting sales for enterprise but this is disclosed.

P2-E Semantic Precision — 4/5
Specific numeric claims throughout: 21,337 servers, 1,782 hosted connectors, 50,000+ businesses, exact per-unit pricing. Avoids vague superlatives.

P3-A Verifiable Performance — 3/5
Public status page at status.glama.ai shows real-time uptime (96.08% over 60 days) with incident timestamps. Self-reported but publicly visible with detailed incident history. No third-party audit found.

P3-B Scoped Permissions — 4/5
Explicit per-tool access control: "Enable/disable specific tools per connector" is a core feature. Managed OAuth credentials with automatic token rotation. Usage analytics by tool. Meaningful agent-scoped permissions.

P3-C Audit Trail — 4/5
Full call logging with inputs and outputs. 100k logs/month per tier. Machine-accessible log data with tiered retention (30/90/180 days). Configurable and tied to agent workspace.

P3-D Behavioral Consistency — 2/5
No API versioning strategy documented. No explicit change notice policy. Terms of service exist but standard boilerplate. No version-controlled API contract or migration policy found.

P4-A Friction-Free Activation — 4/5
Self-serve signup. Registry and MCP Inspector are free with no registration required. Starter plan at $9/mo with immediate API key issuance implied. No sales call gate. Multi-step configuration does add some friction.

P4-B Agent Decision Signals — 3/5
Free registry tier provides immediate value without commitment. Clear pricing ladder ($9→$26→$80). OpenAPI spec available for programmatic evaluation. No explicit trial-before-buy for gateway features.

P5-A Integration Depth — 3/5
Platform manages OAuth tokens, credentials, and MCP server configurations per workspace. Switching away requires reconfiguring all agent integrations and forfeiting log history. Moderate switching cost.

P5-B Agent Memory Layer — 2/5
Usage analytics tracks which tools agents call over time. Team workspaces persist configurations. But no true agent-accessible memory layer — no API endpoint for agents to read/write contextual history.

P5-C Programmatic Renewal — 1/5
Standard subscription billing only. No API endpoint for agents to query subscription status, trigger renewal, or receive expiry signals programmatically.

P5-D Compounding Value — 2/5
Registry grows over time (21k+ servers implies network effects). Agent usage data accumulates in logs. But no explicit agent-readable signal that the platform's value is increasing — network effects are invisible to agents.

Rubric v1 (April 2026). Scores reflect the company's state on the audit date and may have improved since.