Shared risk memory for agents
Give computer-using agents outside-in risk context and inherited conclusions that survive the chat.
Syndu gives computer-using agents outside-in risk context and a governed memory loop. This index maps the supporting docs: plugin install, MCP workflows, the Risk API, pricing, privacy, and the evidence depth behind the system.
Give computer-using agents outside-in risk context and inherited conclusions that survive the chat.
Start with the install surface that puts agents directly into Syndu's governed memory loop.
Use reports for readable evidence and the Risk API for scoring and writeback without changing the underlying workspace memory model.
The clearest entry is the agent workflow itself: investigate something risky, decide what matters, preserve the approved conclusion, and let the next run inherit the work.
The plugin and MCP docs should be the first stop for buyers. Reports and APIs remain essential, but they now support the same shared risk memory system instead of acting like separate product lines.
Reports prove depth and give humans readable evidence. The Risk API gives applications a narrow scoring and writeback contract. Both stay important, but neither should outrank the agent operating loop.
The commercial center of Syndu is not a raw score. It is the governed memory loop that lets one operator or agent resolve a case once and lets future runs start ahead.
The strategic story: why Syndu exists as shared risk memory for agents, where reports and APIs fit, and why inherited conclusions matter.
The `score` endpoint contract, request and response format, validity boundaries, and entitlement visibility.
How customers provision keys, manage API quota, and activate the API subscription service through the workspace control plane.
The hosted MCP server for Codex, Claude Code, Manus, and other agents: indicator resolution, report snapshots, risk explanation, and governed outcome memory with explicit workspace and hive-mind controls.
Download the official Codex plugin from Syndu Web LTD and give Codex shared risk memory, governed workspace context, and inherited conclusions across future runs.
The first workflows to buy and expand: investigate IPs, domains, and orgs, retain handoff memory, draft detections, and extend into fraud or account-abuse investigations.
How the same evidence graph, Risk API, and memory loop support pre-transaction enrichment, manual review, dispute evidence, and merchant-risk operations.
The public catalog of behavioral annotators that explain how Syndu interprets observed activity.
Syndu exposes a dimensional risk surface across network entities and geography: ip address, subnet, ISP, ASN, organization, city, region, and country. That same dimensional graph powers the Risk API, MCP tools, and the memory loop that preserves the resulting decisions.
Syndu is an external context layer for monitoring, hygiene review, agent workflows, and decision support. It complements SoC and SIEM processes rather than replacing them, while sharing the same evidence and memory model used elsewhere on the platform.
Syndu is also a decision surface for fraud, payments, merchant-risk, and trust workflows. The same outside-in evidence graph, Risk API, and governed memory loop help teams score, review, and preserve conclusions around risky identifiers.
The platform supports three access surfaces at scale: direct report access for large-scale browser or agent consumption, the Risk API for application-native scoring, and the MCP server for collaborative agent workflows.
Syndu is one platform with three access surfaces: direct access for report-shell consumption, API access for application-side scoring, and MCP access for agent collaboration. They sit under the same workspace, quota, and subscription control plane, but differ in delivery contract.
Score risky IPs in real time, report lightweight business feedback, and push structured identifier outcomes back into workspace memory from the same authenticated API surface. Syndu keeps the contract simple: `score` is the metered read path, while `rate` and `report` are authenticated writeback endpoints that do not consume score quota. The score response now also carries a live community-signal block so applications can see whether analysts are converging on the same entity right now.
Registration unlocks the workspace control plane, plan selection, credential provisioning, and the subscription flow across web, API, and MCP access.
