Modernise without the mess
Map a tangled legacy system before changing a line. Refactor with the impact in view, not with crossed fingers.
More room to build.
The control you need.
Give developers a frontier coding tool. Give the organisation one place to decide what every agent may do, which models it uses and what it costs, wherever the product runs.
What engineering leaders ask usEnterprise challenges,
Enterprise challenges,
solved.
The things engineering leaders bring to us most often, and what SyphaCode does about each of them inside your own perimeter.
Fix vulnerabilities safely
Patch security gaps without destabilising what already works: context-aware fixes, not blind patches.
Tests that match your code
Unit and integration tests built around your real patterns, not generic templates.
Docs that do not drift
Documentation kept current as the code evolves. No more out-of-date READMEs.
From stack trace to solution
Errors analysed with the whole codebase in context, so production issues resolve faster with fewer guesses.
Productive on day one
New engineers find their way around unfamiliar code immediately. Ramp time drops; contribution time rises.
For the whole teamMore builders.
More builders.
A clearer view.
Give developers room to work. Decide what their agents may do, give every team its specialists and its budget, and follow every run, in one dashboard.
Discuss your team’s requirementsTeams dashboard, illustrative
SyphaCode TeamsExample
SyphaCode TeamsExampleSet the rules once.Agent settings · each level below can only make them stricter
Organisation/Engineering/Contractors
Tool permissionsNot set here: using EngineeringSet at this level
ReadRead filesAllowAskDenyEditCreate or change filesAllowAskDenyBashOrganisation set AskAllowAskDeny- Lock for membersRead-only in their editor and CLI
Applies to 12 membersSaveSaved
A member’s editorLocked by Contractors
BashAskDenyGive every team its specialists.Team agents · written once, given to chosen teams
@code-reviewer
Reviews every change for bugs, style and missing tests.
Published to 2 teamsEngineeringCodePlan@code-reviewer
PlatformCodePlan@code-reviewer
ContractorsCodePlanNot given
See every step of a run.Agent runs · members see their own
OrchestratorAdd CSV export to reports
- PlanMapped the report pipelineDone
- CodeEdited 4 filesDone
- DebugRate limit, retriedRetryingRecovered
Completed · 3 tasks · 1 issue recovered
Know when a check stops something.Security alerts · for admins and Team Managers
CodeA member’s editor
Stopped by a safety check
Last 7 days1 file write stopped1 prompt and 1 file write stopped
Prompt stoppedjust nowA member of EngineeringBulk data extraction
Know where the work goes.Requests through Sypha Gateway
Requests1,248Tokens2.4MAverage response1.8s
Give every team a budget.Team wallet
10,000team credits
Given to members 4,000Left in the wallet 6,000
Analytics covers traffic routed through Sypha Gateway. Direct BYOK requests are outside that reporting path.
Enterprise ControlThe details behind
The details behind
a well-run team.
Configured once in the dashboard, enforced across every IDE and the CLI. The same controls run in every deployment mode.
Agent settings
Decide what agents may do for the whole organisation, a team, a role or one member. Each level can only tighten the one above, and locked settings are read-only in every editor and the CLI.
Team agents
Publish specialist agents to everyone or to chosen teams. Pin their model, limit their tools, and decide whether members may replace them.
Safety alerts
Checks in the editor stop risky prompts and malicious file writes, and alert admins and Team Managers. Only a short excerpt is kept, with secrets removed.
Agent runs
Follow every prompt and the subagents it started. When a run fails, see where, why and what to do next, and export the report.
Team budgets
Fund each team from the organisation wallet. Team Managers give credits to members, and a daily cap can apply per member or per role.
Usage analytics
Cost, prompts and tokens by member, team, model and project, with success rates for agents, skills and tools. Export any view to CSV.
Security and sovereigntyGoverned by
Governed by
architecture, not clauses.
Three principles run through every deployment mode. Cloud tools address these concerns contractually; on premises, the data flow itself is removed.
Capability is a permission
Every tool is Allow, Ask or Deny, with exceptions for paths and commands, decided by the organisation and only tightened below it. Commands can run in an operating-system sandbox, and when policy cannot be fetched, access is denied.
The data path ends inside your perimeter
On premises, prompts, code context, the local index, telemetry and the audit trail all terminate on your infrastructure. No Sypha cloud dependency, no provider key in the serving path, no outbound requirement.
Every action lands in one trail
Requests, tool calls, agent runs, model choices and every change to agent settings are recorded where you decide, per member, exportable for review. The same data is your adoption and cost evidence.
Statements about the on-premises mode describe its architecture. Sypha Cloud and private cloud deployments carry the same controls with a different data path.
Where it runsOne product.
One product.
Three places to run it.
The capability never depends on where it runs. Start in Sypha Cloud, move to your cloud or your own data centre later, and keep the same product, controls and workflows.
Start building. We handle the setup.
A managed service with two routes to inference: open-weight models hosted in India, where Auto Mode can pick the model for each request, or 400+ models such as Claude, GPT and Gemini called through their providers’ own APIs via the Sypha gateway. We do not host those; we route to them. Nothing to run.
- Managed service
- Auto Mode on India-hosted models
- Claude, GPT, Gemini and 400+ more via provider APIs
The cloud your organisation already trusts.
Deploy on Vertex AI, Bedrock, Azure or OCI under your own contracts, self-host open models on cloud GPUs, or run the application layer in a sovereign Indian data centre.
- Your tenant, your keys
- Approved model endpoints
- Regional residency
Nothing leaves the perimeter.
The full stack on your own GPUs behind your own perimeter, with air-gapped options. Self-hosted open-source models only, no outbound calls, and the audit trail on your storage.
- Air-gapped option
- No per-token cost
- Audit on your storage
Hybrid, under one console: regulated production on premises, experimentation in Sypha Cloud, the same controls throughout.
Discuss your setupYour code. Your models.
Your rules.
A pilot on a representative codebase, sizing for your team, and the governance configured before anyone writes a prompt.
Questions, answeredA little more
A little more
clarity.
What does the Teams dashboard show?
Cost, requests, tokens and active members by product, team, member and date; every prompt, broken down by person, model or cost; how often each agent, skill and tool runs and how reliably; model and provider comparisons, cost per project and the lines each change added and removed. Everything exports to CSV. Hosted reporting covers traffic routed through Sypha, not direct BYOK traffic.
Teams and governanceWho can see a team member’s prompts?
Members see their own prompts and runs. A Team Manager sees their team’s, and organisation admins see the whole organisation. Safety alerts go to admins and the member’s Team Manager.
Teams and governanceHow do seats and credits work?
Owners add seats and invite members. The organisation wallet funds a budget for each team, Team Managers allocate credits to their members and can ask for more, and unused allocations return to the pool. Every movement is logged. Seats are billed per active member and prorated.
Teams and governanceCan administrators restrict actions?
Yes. Agent settings cover tool permissions, providers, custom providers, web tools, the sandbox, MCP servers and codebase indexing, for the organisation, a team, a scoped role or one member. Each level can only make the one above stricter, and a locked section cannot be changed below it.
Teams and governanceCan we set spending and token limits?
Yes. Team budgets and a daily credit cap per member or scoped role limit spend, and a session cost alert tells a member when a session passes an amount you set. Reporting covers traffic routed through Sypha; direct BYOK traffic sits outside it.
Teams and governanceIs single sign-on available?
Yes, on Enterprise. Connect your identity provider over OpenID Connect or SAML 2.0, verify your email domains, and require single sign-on for everyone who signs in from them.
Teams and governanceWhat audit evidence exists?
Every request is recorded with the user, model, tokens, tool calls and response time. Every change to agent settings keeps a history of who changed what and when, and Agent Runs keeps each run’s steps and failures. On premises, the audit trail is written to storage you control.
Teams and governanceWhere does our code and usage data go?
It depends on the deployment. In Sypha Cloud, requests go to the managed service; if you pick a provider model such as Claude, GPT or Gemini, the gateway sends the request on to that provider’s API, since those models are not hosted by us. In a private cloud requests stay in your tenant. On premises, nothing leaves your network. With your own keys, requests go straight from your machine to your provider.
Deployment and dataCan SyphaCode run in our own environment?
Yes. Private cloud deployments run on Vertex AI, Bedrock, Azure or OCI under your contracts, or in a sovereign Indian data centre. On-premises deployments run the full stack on your GPUs, with an air-gapped option using self-hosted open-source models.
Deployment and dataDoes an on-premises deployment need internet access?
No. There is no Sypha cloud dependency and no outbound requirement; the extensions and the CLI talk only to your own endpoints. Cloud models, public web access and online integrations are not available in that mode.
Deployment and dataIs our code used to train models?
On premises and in a private cloud, nothing reaches SYPHA, so there is nothing to train on, and self-hosted weights are static files that do not learn from your usage. In Sypha Cloud, administrators can switch off any use of the team’s code for model refinement from the dashboard.
Deployment and dataWhat about data residency obligations?
On premises, no code or prompt crosses a border, so localisation expectations such as India’s DPDP Act are addressed by architecture rather than by contract clause. Private cloud deployments keep data in the region and tenant you choose.
Deployment and dataWhat infrastructure do we need on premises?
It depends on developers, concurrency and the model you choose. Current open-weight coding models run on a single GPU server for a pilot, and rented GPU capacity is an option. We size it with you before anything is installed.
Deployment and dataCan we run more than one mode at once?
Yes. Regulated production work can stay on premises while experimentation runs in Sypha Cloud, under one console with the same controls throughout. Moving between modes needs no migration.
Deployment and dataWhat help is available for a private rollout?
Start with your codebase, team size, preferred models and governance requirements. The engagement can include a pilot on a representative codebase, infrastructure sizing, configuration, onboarding and developer enablement, typically within a few weeks.
Deployment and dataHow do I get help?
The documentation covers setup, agents, the CLI and team administration. Use the contact page for product support; Enterprise plans include priority support with SLAs.
SupportWho makes SyphaCode?
SyphaCode is built, supported and delivered by SYPHA (Sypha AI Private Limited).
Support