Single model gateway
Point agents at one Cohesor endpoint using Anthropic- or OpenAI-compatible protocols. The documented setup uses environment variables and gateway keys, allowing supported clients to connect without SDK changes.
Cohesor
Freemium
Cohesor is an AI gateway and control plane for routing model requests, compressing agent context, governing spend, and brokering MCP tool calls. It is designed for teams running internal or customer-facing AI agents.
Cohesorとは？
Cohesor is an AI gateway and control plane for teams running model-backed agents. It places one endpoint between agents and model providers, then applies policy checks, token compression, model routing, spend controls, and request tracing before forwarding calls.
Agents can connect through Anthropic- or OpenAI-compatible interfaces, including Claude Code, Codex, and compatible SDK clients, without an SDK rewrite. Cohesor's routing layer can balance quality, latency, and cost, while its compression layer reduces prompts and tool outputs before they are sent to a model.
The platform also includes an MCP gateway for governed tool access. Teams can connect supported services such as GitHub, Gmail, Google Drive, and Google Calendar, assign per-agent scopes, broker OAuth credentials, and retain tool-call traces. Usage is attributed by user, team, and API key, with budget caps and billing reports for operational oversight.
Cohesorでできること
Token compression
Compress prompts, conversations, and MCP tool outputs before they re-enter the model context. The documented methods include stale-read removal, context compaction, semantic deduplication, and structural normalization.
Smart and fixed routing
Route requests according to quality, cost, latency, or a configured objective. Cohesor can select among available models, while the pricing FAQ states that users can also pin a single model in fixed mode.
Spend governance and attribution
Apply gateway-key and team scopes, rate limits, and budget guards before requests reach a provider. Track usage and spend by user, team, and API key, with hard caps, showback reports, and billing history.
Governed MCP gateway
Expose enabled MCP connectors through one audited endpoint. Per-agent tool scopes are enforced at the wire, and OAuth tokens are brokered through an encrypted vault rather than placed in prompts or logs.
Request and tool tracing
Record model and tool activity in parent request traces, including usage, latency, cost, routing information, and tool-call arguments. The MCP pages state that agent decisions can be replayed from these traces.
利用シーン
“AI coding assistants”
Route Claude Code, Codex, Cursor, internal development agents, or CI bots through one gateway. Repeated context and tool results can be compressed, while straightforward tasks can use economy models and more demanding work can use stronger models.
“Knowledge and research assistants”
Use Cohesor in enterprise search, RAG, research, or analyst workflows where retrieved documents may be much larger than the user's question. Compression reduces the context passed back to the model, and policy and spend data remain attached to the request.
“Support and customer-facing AI”
Place support agents, in-app chat, checkout, booking, onboarding, concierge, or voice workflows behind the same control plane. Internal users can be attributed by seat, while product usage can be tracked per end user.
“Back-office process automation”
Govern document processing, ticket triage, claims workflows, and other operational agents with budgets, routing policies, and traces. MCP scopes can limit which repositories, mailboxes, calendars, or files a particular agent can access.
“Team-wide model operations”
Give multiple teams a shared gateway for model and tool access while retaining per-user, team, and API-key attribution. The usage-based service includes a pay-as-you-go option and an enterprise tier with additional organization and support features.
よくある質問
How do I connect an agent to Cohesor?
Create a gateway key in the dashboard, then set the relevant Anthropic or OpenAI base URL and API-key environment variables. The quickstart documents Cohesor endpoints for Claude Code and Codex/OpenAI clients and says no SDK or proxy configuration changes are required.
Can I use Cohesor with a fixed model?
Yes. Cohesor can route requests according to configured objectives, and its pricing FAQ says users can pin a single model in fixed mode when they do not want dynamic routing.
Which MCP tools can Cohesor manage?
The MCP pages document GitHub, Gmail, Google Drive, and Google Calendar connectors, and describe the gateway as compatible with MCP tools more broadly. Available connector actions and scopes depend on the enabled integration.
How are MCP credentials handled?
OAuth tokens are encrypted at rest, brokered at call time, and refreshed automatically according to the MCP documentation. The credentials are not placed in prompts, logs, or the agent's context window.
How is Cohesor priced?
The pricing page describes usage-based billing: LLM usage is charged at provider cost plus a flat $10 per 100,000 requests. Enterprise customers can contact sales for custom volume pricing and terms.
クイック情報
- Product type
- AI gateway and control plane
- Primary users
- Teams running internal or customer-facing AI agents
- Model interfaces
- Anthropic- and OpenAI-compatible clients
- Tool protocol
- MCP 1.0 gateway with OAuth 2.1 documentation
- Documented connectors
- GitHub, Gmail, Google Drive, and Google Calendar
- Pricing model
- LLM usage at provider cost plus $10 per 100,000 requests
Cohesor のトラフィック分析
トラフィックデータは参考情報としてご利用ください。
- ドメイン評価
- 1
Cohesorの代替品
Edgee
www.edgee.ai
Edgee is an agent gateway for coding teams that reduces token usage, routes requests across models, and provides usage visibility. Its Compression V2 works as a drop-in CLI layer for Claude Code, Codex, OpenCode, Cursor, and other supported agents.
OneCLI
onecli.sh
OneCLI is an open-source platform for giving employees personal AI agents that can work across company tools. It provides isolated sandboxes, gateway-enforced policies, scoped credentials, approvals, and audit logs for team use.
Docker MCP Enterprise Gateway
www.docker.com
Docker MCP Enterprise Gateway governs how MCP-compatible clients access approved servers and tools. It applies identity-aware policy, supplies credentials from an approved secret store, and records tool-call decisions for enterprise teams.
Agent Router
theagentrouter.ai
Agent Router is an open-source Go project for managing unified access to generative AI services through Envoy Gateway. It is intended for teams that want a gateway-based control point for AI service traffic and related agent or MCP workloads.
treg.to
treg.to
treg.to is an API gateway and MCP server for calling tools and data providers through one endpoint and credential. It helps agents and developers use cataloged SEO, enrichment, advertising, social, scraping, and other data tools without managing separate provider keys for every metered call.
Obot
obot.ai
Obot is an open-source enterprise AI control plane and MCP gateway for deploying, governing, and auditing AI agents, MCP servers, and Skills. It helps organizations centralize access policies, discovery, authentication, and activity records across AI tooling.